open-mmlab / mmdetection
run from
| BOX AP | AP50 | AP75 | APS | APM | APL |
SPEED
|
|||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MODEL | CODE | PAPER |
ε-REPR
|
CODE | PAPER |
ε-REPR
|
CODE | PAPER |
ε-REPR
|
CODE | PAPER |
ε-REPR
|
CODE | PAPER |
ε-REPR
|
CODE | PAPER |
ε-REPR
|
PAPER |
GLOBAL RANK
|
|||||||||
|
Cascade Mask R-CNN
(ResNet-101-FPN, 1x LR) |
0.426 | -- | 0.607 | -- | 0.467 | -- | 0.238 | -- | 0.464 | -- | 0.569 | -- | 7.5 | #64 | |||||||||||||||
|
Cascade Mask R-CNN
(ResNet-101-FPN, 20e LR) |
0.433 | 0.433 | 0.613 | 0.617 | 0.470 | 0.472 | 0.244 | 0.242 | 0.469 | 0.463 |
|
0.580 | 0.582 | 7.7 | #24 | ||||||||||||||
|
Cascade Mask R-CNN
(ResNet-50-FPN, 1x LR) |
0.412 | 0.413 | 0.591 | 0.594 | 0.451 | 0.453 | 0.233 | 0.232 | 0.445 | 0.438 |
|
0.545 | 0.558 | 8.1 | #47 | ||||||||||||||
|
Cascade Mask R-CNN
(ResNet-50-FPN, 20e LR) |
0.423 | -- | 0.605 | -- | 0.460 | -- | 0.237 | -- | 0.457 | -- | 0.564 | -- | 8.4 | #37 | |||||||||||||||
|
Cascade Mask R-CNN
(ResNeXt-101 32x4d-FPN, 1x LR) |
0.444 | -- | 0.626 | -- | 0.486 | -- | 0.254 | -- | 0.481 | -- | 0.587 | -- | 6.8 | #22 | |||||||||||||||
|
Cascade Mask R-CNN
(ResNeXt-101 32x4d-FPN, 20e LR) |
0.447 | -- | 0.630 | -- | 0.489 | -- | 0.259 | -- | 0.487 | -- | 0.589 | -- | 6.9 | #16 | |||||||||||||||
|
Cascade Mask R-CNN
(ResNeXt-101 64x4d-FPN, 1x LR) |
0.452 | -- | 0.639 | -- | 0.492 | -- | 0.262 | -- | 0.496 | -- | 0.596 | -- | 5.6 | #13 | |||||||||||||||
|
Cascade Mask R-CNN
(ResNeXt-101 64x4d-FPN, 20e LR) |
0.457 | 0.458 | 0.641 | 0.641 | 0.500 | 0.503 | 0.262 | 0.272 | 0.496 | 0.495 | 0.600 | 0.601 | 5.7 | #16 | |||||||||||||||
|
Cascade R-CNN
(HRNetV2p-W32) |
0.436 | -- | 0.615 | -- | 0.476 | -- | 0.255 | -- | 0.464 | -- | 0.573 | -- | 8.7 | #26 | |||||||||||||||
|
Cascade R-CNN
(ResNet-101-FPN, 1x LR) |
0.420 | -- | 0.603 | -- | 0.459 | -- | 0.232 | -- | 0.459 | -- | 0.563 | -- | 10.6 | #40 | |||||||||||||||
|
Cascade R-CNN
(ResNet-101-FPN+, cascade) |
0.425 | -- | 0.607 | -- | 0.463 | -- | 0.237 | -- | 0.461 | -- | 0.569 | -- | 10.7 | #39 | |||||||||||||||
|
Cascade R-CNN
(ResNet-50-FPN+) |
0.404 | -- | 0.585 | -- | 0.439 | -- | 0.215 | -- | 0.437 | -- | 0.538 | -- | 12.1 | #88 | |||||||||||||||
|
Cascade R-CNN
(ResNet-50-FPN, 20e LR) |
0.411 | -- | 0.591 | -- | 0.448 | -- | 0.225 | -- | 0.444 | -- | 0.549 | -- | 12.3 | #50 | |||||||||||||||
|
Cascade R-CNN
(ResNeXt-101 32x4d-FPN, 1x LR) |
0.436 | -- | 0.622 | -- | 0.474 | -- | 0.250 | -- | 0.477 | -- | 0.574 | -- | 9.1 | #25 | |||||||||||||||
|
Cascade R-CNN
(ResNeXt-101 32x4d-FPN, 20e LR) |
0.440 | -- | 0.625 | -- | 0.480 | -- | 0.253 | -- | 0.478 | -- | 0.581 | -- | 9.2 | #29 | |||||||||||||||
|
Cascade R-CNN
(ResNeXt-101 64x4d-FPN, 1x LR) |
0.445 | -- | 0.633 | -- | 0.486 | -- | 0.261 | -- | 0.481 | -- | 0.591 | -- | 7.0 | #14 | |||||||||||||||
|
Cascade R-CNN
(ResNeXt-101 64x4d-FPN, 20e LR) |
0.447 | -- | 0.631 | -- | 0.490 | -- | 0.258 | -- | 0.483 | -- | 0.588 | -- | 7.0 | #18 | |||||||||||||||
|
Faster R-CNN
(HRNetV2p-W18) |
0.383 | -- | 0.593 | -- | 0.417 | -- | 0.223 | -- | 0.414 | -- | 0.491 | -- | 9.3 | #80 | |||||||||||||||
|
Faster R-CNN
(HRNetV2p-W18, 1x LR) |
0.361 | 0.362 | 0.573 | 0.573 | 0.389 | 0.393 | 0.220 | 0.207 |
|
0.389 | 0.390 | 0.469 | 0.468 | 9.1 | #100 | ||||||||||||||
|
Faster R-CNN
(HRNetV2p-W32) |
0.406 | -- | 0.619 | -- | 0.441 | -- | 0.239 | -- | 0.441 | -- | 0.524 | -- | 9.6 | #55 | |||||||||||||||
|
Faster R-CNN
(HRNetV2p-W32, 1x LR) |
0.395 | 0.396 | 0.610 | 0.610 | 0.431 | 0.433 | 0.236 | 0.237 | 0.429 | 0.425 | 0.510 | 0.505 | 9.2 | #64 | |||||||||||||||
|
Faster R-CNN
(ResNet-101-FPN, 1x LR) |
0.385 | -- | 0.603 | -- | 0.416 | -- | 0.223 | -- | 0.430 | -- | 0.498 | -- | 11.8 | #86 | |||||||||||||||
|
Faster R-CNN
(ResNet-101-FPN, 2x LR) |
0.394 | -- | 0.606 | -- | 0.430 | -- | 0.221 | -- | 0.436 | -- | 0.521 | -- | 11.9 | #65 | |||||||||||||||
|
Faster R-CNN
(ResNet-50-FPN, 1x LR) |
0.364 | -- | 0.584 | -- | 0.391 | -- | 0.215 | -- | 0.400 | -- | 0.466 | -- | 13.8 | #100 | |||||||||||||||
|
Faster R-CNN
(ResNet-50-FPN, 2x LR) |
0.377 | -- | 0.592 | -- | 0.411 | -- | 0.219 | -- | 0.414 | -- | 0.487 | -- | 14.0 | #83 | |||||||||||||||
|
Faster R-CNN
(ResNet-50-FPN, GN+WS, LR 1x) |
0.389 | -- | 0.604 | -- | 0.421 | -- | 0.235 | -- | 0.424 | -- | 0.504 | -- | 9.9 | #67 | |||||||||||||||
|
Faster R-CNN
(ResNeXt-101 32x4d-FPN, 1x LR) |
0.401 | -- | 0.620 | -- | 0.438 | -- | 0.234 | -- | 0.446 | -- | 0.517 | -- | 10.1 | #41 | |||||||||||||||
|
Faster R-CNN
(ResNeXt-101 32x4d-FPN, 2x LR) |
0.404 | -- | 0.619 | -- | 0.441 | -- | 0.233 | -- | 0.446 | -- | 0.529 | -- | 10.2 | #44 | |||||||||||||||
|
Faster R-CNN
(ResNeXt-101 64x4d-FPN, 1x LR) |
0.413 | -- | 0.633 | -- | 0.452 | -- | 0.244 | -- | 0.458 | -- | 0.534 | -- | 7.5 | #26 | |||||||||||||||
|
Faster R-CNN
(ResNeXt-101 64x4d-FPN, 2x LR) |
0.407 | -- | 0.620 | -- | 0.446 | -- | 0.229 | -- | 0.445 | -- | 0.536 | -- | 7.6 | #53 | |||||||||||||||
|
HTC
(HRNetV2p-W32) |
0.453 | -- | 0.636 | -- | 0.491 | -- | 0.270 | -- | 0.484 | -- | 0.595 | -- | 3.8 | #17 | |||||||||||||||
|
HTC
(ResNet-101-FPN) |
0.449 | -- | 0.638 | -- | 0.487 | -- | 0.264 | -- | 0.483 | -- | 0.599 | -- | 4.2 | #17 | |||||||||||||||
|
HTC
(ResNet-50-FPN, 1x LR) |
0.421 | -- | 0.608 | -- | 0.459 | -- | 0.239 | -- | 0.455 | -- | 0.562 | -- | 3.6 | #49 | |||||||||||||||
|
HTC
(ResNet-50-FPN, 20e LR) |
0.432 | -- | 0.621 | -- | 0.468 | -- | 0.249 | -- | 0.464 | -- | 0.578 | -- | 4.3 | #36 | |||||||||||||||
|
HTC
(ResNeXt-101 32x4d-FPN) |
0.461 | -- | 0.651 | -- | 0.502 | -- | 0.275 | -- | 0.498 | -- | 0.612 | -- | 3.9 | #12 | |||||||||||||||
|
HTC
(ResNeXt-101 64x4d-FPN) |
0.469 | -- | 0.660 | -- | 0.512 | -- | 0.280 | -- | 0.507 | -- | 0.621 | -- | 3.5 | #7 | |||||||||||||||
|
HTC
(ResNeXt-101 64x4d-FPN, DCN, multi-scale) |
0.506 | -- | 0.701 | -- | 0.551 | -- | 0.328 | -- | 0.547 | -- | 0.664 | -- | 3.2 | #1 | |||||||||||||||
|
Libra R-CNN
(Faster R-CNN, ResNet-101-FPN) |
0.403 | -- | 0.612 | -- | 0.439 | -- | 0.233 | -- | 0.443 | -- | 0.522 | -- | #58 | ||||||||||||||||
|
Libra R-CNN
(Faster R-CNN, ResNet-50-FPN) |
0.385 | -- | 0.595 | -- | 0.425 | -- | 0.229 | -- | 0.418 | -- | 0.489 | -- | #67 | ||||||||||||||||
|
Libra R-CNN
(Faster R-CNN, ResNeXt-101 64x4d-FPN) |
0.427 | -- | 0.638 | -- | 0.468 | -- | 0.258 | -- | 0.466 | -- | 0.554 | -- | #30 | ||||||||||||||||
|
Libra R-CNN
(RetinaNet, ResNet-50-FPN) |
0.377 | -- | 0.570 | -- | 0.402 | -- | 0.216 | -- | 0.417 | -- | 0.495 | -- | #96 | ||||||||||||||||
|
Mask R-CNN
(HRNetV2p-W18) |
0.392 | -- | 0.601 | -- | 0.429 | -- | 0.241 | -- | 0.422 | -- | 0.507 | -- | 7.2 | #71 | |||||||||||||||
|
Mask R-CNN
(HRNetV2p-W18, 1x LR) |
0.372 | 0.371 | 0.581 | -- | 0.406 | -- | 0.221 | 0.219 | 0.403 | 0.395 |
|
0.476 | 0.479 | 7.0 | #94 | ||||||||||||||
|
Mask R-CNN
(HRNetV2p-W32) |
0.417 | -- | 0.624 | -- | 0.457 | -- | 0.251 | -- | 0.453 | -- | 0.543 | -- | 7.6 | #47 | |||||||||||||||
|
Mask R-CNN
(HRNetV2p-W32, 1x LR) |
0.407 | 0.409 | 0.618 | -- | 0.447 | -- | 0.252 | 0.245 |
|
0.444 | 0.439 | 0.518 | 0.522 | 7.3 | #66 | ||||||||||||||
|
Mask R-CNN
(HRNetV2p-W32, cascade) |
0.445 | -- | 0.623 | -- | 0.486 | -- | 0.261 | -- | 0.479 | -- | 0.585 | -- | 6.5 | #21 | |||||||||||||||
|
Mask R-CNN
(ResNet-50-FPN, GN+WS, LR 20-23-24e) |
0.410 | -- | 0.619 | -- | 0.448 | -- | 0.245 | -- | 0.446 | -- | 0.537 | -- | 7.9 | #51 | |||||||||||||||
|
Mask R-CNN
(ResNet-50-FPN, GN+WS, LR 2x) |
0.402 | -- | 0.608 | -- | 0.435 | -- | 0.235 | -- | 0.433 | -- | 0.534 | -- | 8.2 | #74 | |||||||||||||||
|
Mask R-CNN
(ResNeXt-101 32x4d-FPN, GN+WS, LR 1x) |
0.433 | -- | 0.641 | -- | 0.477 | -- | 0.252 | -- | 0.473 | -- | 0.569 | -- | 5.9 | #25 | |||||||||||||||
|
Mask Scoring R-CNN
(ResNet-101-FPN, LR 1x) |
0.398 | -- | 0.612 | -- | 0.434 | -- | 0.233 | -- | 0.439 | -- | 0.514 | -- | #66 | ||||||||||||||||
|
Mask Scoring R-CNN
(ResNet-101-FPN, LR 2x) |
0.407 | -- | 0.617 | -- | 0.446 | -- | 0.236 | -- | 0.446 | -- | 0.536 | -- | #55 | ||||||||||||||||
|
Mask Scoring R-CNN
(ResNet-50-FPN, LR 1x) |
0.374 | -- | 0.590 | -- | 0.405 | -- | 0.214 | -- | 0.409 | -- | 0.491 | -- | #91 | ||||||||||||||||
|
Mask Scoring R-CNN
(ResNet-50-FPN, LR 2x) |
0.382 | -- | 0.593 | -- | 0.414 | -- | 0.214 | -- | 0.414 | -- | 0.513 | -- | #80 | ||||||||||||||||
|
Mask Scoring R-CNN
(ResNeXt-101 64x4d-FPN, LR 1x) |
0.420 | -- | 0.640 | -- | 0.460 | -- | 0.248 | -- | 0.465 | -- | 0.543 | -- | #18 | ||||||||||||||||
|
Mask Scoring R-CNN
(ResNeXt-101 64x4d-FPN, LR 2x) |
0.422 | -- | 0.634 | -- | 0.461 | -- | 0.242 | -- | 0.460 | -- | 0.561 | -- | #37 | ||||||||||||||||
|
RetinaNet
(ResNet-101-FPN, 1x LR) |
0.377 | -- | 0.575 | -- | 0.404 | -- | 0.211 | -- | 0.422 | -- | 0.495 | -- | 10.1 | #88 | |||||||||||||||
|
RetinaNet
(ResNet-101-FPN+, cascade) |
0.381 | -- | 0.581 | -- | 0.406 | -- | 0.202 | -- | 0.418 | -- | 0.508 | -- | 10.4 | #97 | |||||||||||||||
|
RetinaNet
(ResNet-50-FPN+) |
0.356 | -- | 0.555 | -- | 0.383 | -- | 0.200 | -- | 0.396 | -- | 0.468 | -- | 11.0 | #100 | |||||||||||||||
|
RetinaNet
(ResNet-50-FPN, 20e LR) |
0.364 | -- | 0.563 | -- | 0.387 | -- | 0.193 | -- | 0.399 | -- | 0.489 | -- | 11.9 | #98 | |||||||||||||||
|
RetinaNet
(ResNeXt-101 32x4d-FPN, 1x LR) |
0.390 | -- | 0.594 | -- | 0.417 | -- | 0.226 | -- | 0.434 | -- | 0.509 | -- | 8.9 | #77 | |||||||||||||||
|
RetinaNet
(ResNeXt-101 32x4d-FPN, 20e LR) |
0.393 | -- | 0.598 | -- | 0.423 | -- | 0.210 | -- | 0.436 | -- | 0.523 | -- | 9.1 | #69 | |||||||||||||||
|
RetinaNet
(ResNeXt-101 64x4d-FPN, 1x LR) |
0.400 | -- | 0.609 | -- | 0.430 | -- | 0.235 | -- | 0.444 | -- | 0.526 | -- | 6.9 | #57 | |||||||||||||||
|
RetinaNet
(ResNeXt-101 64x4d-FPN, 20e LR) |
0.396 | -- | 0.603 | -- | 0.423 | -- | 0.216 | -- | 0.435 | -- | 0.535 | -- | 7.0 | #74 | |||||||||||||||
|
SSD300
|
0.257 | -- | 0.439 | -- | 0.262 | -- | 0.069 | -- | 0.277 | -- | 0.426 | -- | 16.8 | #102 | |||||||||||||||
|
SSD512
|
0.293 | -- | 0.492 | -- | 0.308 | -- | 0.118 | -- | 0.341 | -- | 0.447 | -- | 15.0 | #101 | |||||||||||||||
+get badge code
[](https://sotabench.com/user/PartyParrot/repos/deepparrot/mmdetection)
How the Repository is Evaluated
The fullsotabench.py file - source
import argparse
import os
import os.path as osp
import shutil
import tempfile
import urllib.request
import json
from sotabencheval.object_detection import COCOEvaluator
import copy
import mmcv
import torch
import torch.distributed as dist
torch.manual_seed(0)
from mmcv.parallel import MMDataParallel, MMDistributedDataParallel
from mmcv.runner import get_dist_info, load_checkpoint
from mmdet.apis import init_dist
from mmdet.core import coco_eval, wrap_fp16_model
from mmdet.datasets import build_dataloader, build_dataset
from mmdet.models import build_detector
# Extract val2017 zip
from torchbench.utils import extract_archive
image_dir_zip = osp.join('./.data/vision/coco', 'val2017.zip')
extract_archive(from_path=image_dir_zip, to_path='./.data/vision/coco')
from pathlib import Path
if not os.path.isdir(Path.home() / '.cache/torch'):
os.mkdir(Path.home() / '.cache/torch')
def xyxy2xywh(bbox):
_bbox = bbox.tolist()
return [
_bbox[0],
_bbox[1],
_bbox[2] - _bbox[0] + 1,
_bbox[3] - _bbox[1] + 1,
]
def proposal2json(dataset, results):
json_results = []
for idx in range(len(dataset)):
img_id = dataset.img_ids[idx]
bboxes = results[idx]
for i in range(bboxes.shape[0]):
data = dict()
data['image_id'] = img_id
data['bbox'] = xyxy2xywh(bboxes[i])
data['score'] = float(bboxes[i][4])
data['category_id'] = 1
json_results.append(data)
return json_results
def det2json(dataset, results):
json_results = []
for idx in range(len(dataset)):
img_id = dataset.img_ids[idx]
try:
result = results[idx]
except IndexError:
break
for label in range(len(result)):
bboxes = result[label]
for i in range(bboxes.shape[0]):
data = dict()
data['image_id'] = img_id
data['bbox'] = xyxy2xywh(bboxes[i])
data['score'] = float(bboxes[i][4])
data['category_id'] = dataset.cat_ids[label]
json_results.append(data)
return json_results
def segm2json(dataset, results):
bbox_json_results = []
segm_json_results = []
for idx in range(len(dataset)):
img_id = dataset.img_ids[idx]
try:
det, seg = results[idx]
except IndexError:
break
for label in range(len(det)):
# bbox results
bboxes = det[label]
for i in range(bboxes.shape[0]):
data = dict()
data['image_id'] = img_id
data['bbox'] = xyxy2xywh(bboxes[i])
data['score'] = float(bboxes[i][4])
data['category_id'] = dataset.cat_ids[label]
bbox_json_results.append(data)
# segm results
# some detectors use different score for det and segm
if len(seg) == 2:
segms = seg[0][label]
mask_score = seg[1][label]
else:
segms = seg[label]
mask_score = [bbox[4] for bbox in bboxes]
for i in range(bboxes.shape[0]):
data = dict()
data['image_id'] = img_id
data['score'] = float(mask_score[i])
data['category_id'] = dataset.cat_ids[label]
segms[i]['counts'] = segms[i]['counts'].decode()
data['segmentation'] = segms[i]
segm_json_results.append(data)
return bbox_json_results, segm_json_results
def cached_results2json(dataset, results, out_file):
result_files = dict()
if isinstance(results[0], list):
json_results = det2json(dataset, results)
result_files['bbox'] = '{}.{}.json'.format(out_file, 'bbox')
result_files['proposal'] = '{}.{}.json'.format(out_file, 'bbox')
mmcv.dump(json_results, result_files['bbox'])
elif isinstance(results[0], tuple):
json_results = segm2json(dataset, results)
result_files['bbox'] = '{}.{}.json'.format(out_file, 'bbox')
result_files['proposal'] = '{}.{}.json'.format(out_file, 'bbox')
result_files['segm'] = '{}.{}.json'.format(out_file, 'segm')
mmcv.dump(json_results[0], result_files['bbox'])
mmcv.dump(json_results[1], result_files['segm'])
elif isinstance(results[0], np.ndarray):
json_results = proposal2json(dataset, results)
result_files['proposal'] = '{}.{}.json'.format(out_file, 'proposal')
mmcv.dump(json_results, result_files['proposal'])
else:
raise TypeError('invalid type of results')
return result_files
def single_gpu_test(model, data_loader, show=False, evaluator=None):
model.eval()
results = []
dataset = data_loader.dataset
prog_bar = mmcv.ProgressBar(len(dataset))
for i, data in enumerate(data_loader):
with torch.no_grad():
result = model(return_loss=False, rescale=not show, **data)
results.append(result)
if i == 0:
temp_result_files = cached_results2json(copy.deepcopy(dataset), copy.deepcopy(results), 'temp_results.pkl')
anns = json.load(open(temp_result_files['bbox']))
evaluator.add(anns)
from sotabencheval.object_detection.utils import get_coco_metrics
print(evaluator.batch_hash)
print(evaluator.cache_exists)
if evaluator.cache_exists:
return results, True
if show:
model.module.show_result(data, result, dataset.img_norm_cfg)
batch_size = data['img'][0].size(0)
for _ in range(batch_size):
prog_bar.update()
return results, False
def multi_gpu_test(model, data_loader, tmpdir=None):
model.eval()
results = []
dataset = data_loader.dataset
rank, world_size = get_dist_info()
if rank == 0:
prog_bar = mmcv.ProgressBar(len(dataset))
for i, data in enumerate(data_loader):
with torch.no_grad():
result = model(return_loss=False, rescale=True, **data)
results.append(result)
if rank == 0:
batch_size = data['img'][0].size(0)
for _ in range(batch_size * world_size):
prog_bar.update()
# collect results from all ranks
results = collect_results(results, len(dataset), tmpdir)
return results
def collect_results(result_part, size, tmpdir=None):
rank, world_size = get_dist_info()
# create a tmp dir if it is not specified
if tmpdir is None:
MAX_LEN = 512
# 32 is whitespace
dir_tensor = torch.full((MAX_LEN, ),
32,
dtype=torch.uint8,
device='cuda')
if rank == 0:
tmpdir = tempfile.mkdtemp()
tmpdir = torch.tensor(
bytearray(tmpdir.encode()), dtype=torch.uint8, device='cuda')
dir_tensor[:len(tmpdir)] = tmpdir
dist.broadcast(dir_tensor, 0)
tmpdir = dir_tensor.cpu().numpy().tobytes().decode().rstrip()
else:
mmcv.mkdir_or_exist(tmpdir)
# dump the part result to the dir
mmcv.dump(result_part, osp.join(tmpdir, 'part_{}.pkl'.format(rank)))
dist.barrier()
# collect all parts
if rank != 0:
return None
else:
# load results of all parts from tmp dir
part_list = []
for i in range(world_size):
part_file = osp.join(tmpdir, 'part_{}.pkl'.format(i))
part_list.append(mmcv.load(part_file))
# sort the results
ordered_results = []
for res in zip(*part_list):
ordered_results.extend(list(res))
# the dataloader may pad some samples
ordered_results = ordered_results[:size]
# remove tmp dir
shutil.rmtree(tmpdir)
return ordered_results
def evaluate_model(model_name, paper_arxiv_id, weights_url, weights_name, paper_results, config):
print('---')
print('Now Evaluating %s' % model_name)
evaluator = COCOEvaluator(
root='./.data/vision/coco',
model_name=model_name,
paper_arxiv_id=paper_arxiv_id,
paper_results=paper_results)
out = 'results.pkl'
launcher = 'none'
if out is not None and not out.endswith(('.pkl', '.pickle')):
raise ValueError('The output file must be a pkl file.')
cfg = mmcv.Config.fromfile(config)
cfg.data.test['ann_file'] = './.data/vision/coco/annotations/instances_val2017.json'
cfg.data.test['img_prefix'] = './.data/vision/coco/val2017/'
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
cfg.data.test.test_mode = True
# init distributed env first, since logger depends on the dist info.
if launcher == 'none':
distributed = False
else:
distributed = True
init_dist(launcher, **cfg.dist_params)
# build the dataloader
# TODO: support multiple images per gpu (only minor changes are needed)
dataset = build_dataset(cfg.data.test)
data_loader = build_dataloader(
dataset,
imgs_per_gpu=1,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False)
# build the model and load checkpoint
model = build_detector(cfg.model, train_cfg=None, test_cfg=cfg.test_cfg)
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
local_checkpoint, _ = urllib.request.urlretrieve(
weights_url,
'%s/.cache/torch/%s' % (str(Path.home()), weights_name))
print(local_checkpoint)
# '/home/ubuntu/GCNet/mask_rcnn_r50_fpn_1x_20181010-069fa190.pth'
checkpoint = load_checkpoint(model, local_checkpoint, map_location='cpu')
# old versions did not save class info in checkpoints, this walkaround is
# for backward compatibility
if 'CLASSES' in checkpoint['meta']:
model.CLASSES = checkpoint['meta']['CLASSES']
else:
model.CLASSES = dataset.CLASSES
evaluator.reset_time()
if not distributed:
model = MMDataParallel(model, device_ids=[0])
outputs, cache_exists = single_gpu_test(model, data_loader, False, evaluator)
else:
model = MMDistributedDataParallel(model.cuda())
outputs = multi_gpu_test(model, data_loader, args.tmpdir)
if cache_exists:
print('Cache exists: %s' % (evaluator.batch_hash))
evaluator.save()
else:
from mmdet.core import results2json
rank, _ = get_dist_info()
if out and rank == 0:
print('\nwriting results to {}'.format(out))
mmcv.dump(outputs, out)
eval_types = ['bbox']
if eval_types:
print('Starting evaluate {}'.format(' and '.join(eval_types)))
if eval_types == ['proposal_fast']:
result_file = out
else:
if not isinstance(outputs[0], dict):
result_files = results2json(dataset, outputs, out)
else:
for name in outputs[0]:
print('\nEvaluating {}'.format(name))
outputs_ = [out[name] for out in outputs]
result_file = out + '.{}'.format(name)
result_files = results2json(dataset, outputs_,
result_file)
anns = json.load(open(result_files['bbox']))
evaluator.detections = []
evaluator.add(anns)
evaluator.save()
model_configs = []
# libra r-cnn
model_configs.append(
{'model_name': 'Libra R-CNN (Faster R-CNN, ResNet-50-FPN)',
'paper_arxiv_id': '1904.02701',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/libra_rcnn/libra_faster_rcnn_r50_fpn_1x_20190610-bf0ea559.pth',
'weights_name': 'libra_faster_rcnn_r50_fpn_1x_20190610-bf0ea559.pth',
'config': './configs/libra_rcnn/libra_faster_rcnn_r50_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Libra R-CNN (Faster R-CNN, ResNet-101-FPN)',
'paper_arxiv_id': '1904.02701',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/libra_rcnn/libra_faster_rcnn_r101_fpn_1x_20190525-94e94051.pth',
'weights_name': 'libra_faster_rcnn_r101_fpn_1x_20190525-94e94051.pth',
'config': './configs/libra_rcnn/libra_faster_rcnn_r101_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Libra R-CNN (Faster R-CNN, ResNeXt-101 64x4d-FPN)',
'paper_arxiv_id': '1904.02701',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/libra_rcnn/libra_faster_rcnn_x101_64x4d_fpn_1x_20190525-359c134a.pth',
'weights_name': 'libra_faster_rcnn_x101_64x4d_fpn_1x_20190525-359c134a.pth',
'config': './configs/libra_rcnn/libra_faster_rcnn_x101_64x4d_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Libra R-CNN (RetinaNet, ResNet-50-FPN)',
'paper_arxiv_id': '1904.02701',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/libra_rcnn/libra_retinanet_r50_fpn_1x_20190525-ead2a6bb.pth',
'weights_name': 'libra_retinanet_r50_fpn_1x_20190525-ead2a6bb.pth',
'config': './configs/libra_rcnn/libra_retinanet_r50_fpn_1x.py'}
)
# mask scoring r-cnn
model_configs.append(
{'model_name': 'Mask Scoring R-CNN (ResNet-50-FPN, LR 1x)',
'paper_arxiv_id': '1903.00241',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/ms-rcnn/ms_rcnn_r50_caffe_fpn_1x_20190624-619934b5.pth',
'weights_name': 'ms_rcnn_r50_caffe_fpn_1x_20190624-619934b5.pth',
'config': './configs/ms_rcnn/ms_rcnn_r50_caffe_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Mask Scoring R-CNN (ResNet-50-FPN, LR 2x)',
'paper_arxiv_id': '1903.00241',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ms-rcnn/ms_rcnn_r50_caffe_fpn_2x_20190525-a07be31e.pth',
'weights_name': 'ms_rcnn_r50_caffe_fpn_2x_20190525-a07be31e.pth',
'config': './configs/ms_rcnn/ms_rcnn_r50_caffe_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Mask Scoring R-CNN (ResNet-101-FPN, LR 1x)',
'paper_arxiv_id': '1903.00241',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/ms-rcnn/ms_rcnn_r101_caffe_fpn_1x_20190624-677a5548.pth',
'weights_name': 'ms_rcnn_r101_caffe_fpn_1x_20190624-677a5548.pth',
'config': './configs/ms_rcnn/ms_rcnn_r101_caffe_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Mask Scoring R-CNN (ResNet-101-FPN, LR 2x)',
'paper_arxiv_id': '1903.00241',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ms-rcnn/ms_rcnn_r101_caffe_fpn_2x_20190525-4aee1528.pth',
'weights_name': 'ms_rcnn_r101_caffe_fpn_2x_20190525-4aee1528.pth',
'config': './configs/ms_rcnn/ms_rcnn_r101_caffe_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Mask Scoring R-CNN (ResNeXt-101 64x4d-FPN, LR 1x)',
'paper_arxiv_id': '1903.00241',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/ms-rcnn/ms_rcnn_x101_64x4d_fpn_1x_20190628-dec32bda.pth',
'weights_name': 'ms_rcnn_x101_64x4d_fpn_1x_20190628-dec32bda.pth',
'config': './configs/ms_rcnn/ms_rcnn_x101_64x4d_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Mask Scoring R-CNN (ResNeXt-101 64x4d-FPN, LR 2x)',
'paper_arxiv_id': '1903.00241',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ms-rcnn/ms_rcnn_x101_64x4d_fpn_2x_20190525-c044c25a.pth',
'weights_name': 'ms_rcnn_x101_64x4d_fpn_2x_20190525-c044c25a.pth',
'config': './configs/ms_rcnn/ms_rcnn_x101_64x4d_fpn_1x.py'}
)
# Weight Standardization
model_configs.append(
{'model_name': 'Faster R-CNN (ResNet-50-FPN, GN+WS, LR 1x)',
'paper_arxiv_id': '1903.10520',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/faster_rcnn_r50_fpn_gn_ws_1x_20190418-935d00b6.pth',
'weights_name': 'faster_rcnn_r50_fpn_gn_ws_1x_20190418-935d00b6.pth',
'config': './configs/gn+ws/faster_rcnn_r50_fpn_gn_ws_1x.py'}
)
model_configs.append(
{'model_name': 'Mask R-CNN (ResNet-50-FPN, GN+WS, LR 2x)',
'paper_arxiv_id': '1903.10520',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/mask_rcnn_r50_fpn_gn_ws_2x_20190419-9ec97bbb.pth',
'weights_name': 'mask_rcnn_r50_fpn_gn_ws_2x_20190419-9ec97bbb.pth',
'config': './configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_2x.py'}
)
model_configs.append(
{'model_name': 'Mask R-CNN (ResNet-50-FPN, GN+WS, LR 20-23-24e)',
'paper_arxiv_id': '1903.10520',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/mask_rcnn_r50_fpn_gn_ws_20_23_24e_20190425-1d9e499e.pth',
'weights_name': 'mask_rcnn_r50_fpn_gn_ws_20_23_24e_20190425-1d9e499e.pth',
'config': './configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_20_23_24e.py'}
)
model_configs.append(
{'model_name': 'Mask R-CNN (ResNeXt-101 32x4d-FPN, GN+WS, LR 1x)',
'paper_arxiv_id': '1903.10520',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/mask_rcnn_x101_32x4d_fpn_gn_ws_20_23_24e_20190425-1ff3e5b2.pth',
'weights_name': 'mask_rcnn_x101_32x4d_fpn_gn_ws_20_23_24e_20190425-1ff3e5b2.pth',
'config': './configs/gn+ws/mask_rcnn_x101_32x4d_fpn_gn_ws_2x.py'}
)
# SSD
model_configs.append(
{'model_name': 'SSD300',
'paper_arxiv_id': '1512.02325',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ssd300_coco_vgg16_caffe_120e_20181221-84d7110b.pth',
'weights_name': 'ssd300_coco_vgg16_caffe_120e_20181221-84d7110b.pth',
'config': './configs/ssd300_coco.py',
'paper_results': None}
)
model_configs.append(
{'model_name': 'SSD512',
'paper_arxiv_id': '1512.02325',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ssd512_coco_vgg16_caffe_120e_20181221-d48b0be8.pth',
'weights_name': 'ssd512_coco_vgg16_caffe_120e_20181221-d48b0be8.pth',
'config': './configs/ssd512_coco.py',
'paper_results': None}
)
## Hybrid Task Cascade (HTC)
model_configs.append(
{'model_name': 'HTC (ResNet-50-FPN, 1x LR)',
'paper_arxiv_id': '1901.07518',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_r50_fpn_1x_20190408-878c1712.pth',
'weights_name': 'htc_r50_fpn_1x_20190408-878c1712.pth',
'config': './configs/htc/htc_r50_fpn_1x.py',
'paper_results': None}
)
model_configs.append(
{'model_name': 'HTC (ResNet-50-FPN, 20e LR)',
'paper_arxiv_id': '1901.07518',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_r50_fpn_20e_20190408-c03b7015.pth',
'weights_name': 'htc_r50_fpn_20e_20190408-c03b7015.pth',
'config': './configs/htc/htc_r50_fpn_20e.py',
'paper_results': None}
)
model_configs.append(
{'model_name': 'HTC (ResNet-101-FPN)',
'paper_arxiv_id': '1901.07518',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_r101_fpn_20e_20190408-a2e586db.pth',
'weights_name': 'htc_r101_fpn_20e_20190408-a2e586db.pth',
'config': './configs/htc/htc_r101_fpn_20e.py',
'paper_results': None}
)
model_configs.append(
{'model_name': 'HTC (ResNeXt-101 32x4d-FPN)',
'paper_arxiv_id': '1901.07518',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_x101_32x4d_fpn_20e_20190408-9eae4d0b.pth',
'weights_name': 'htc_x101_32x4d_fpn_20e_20190408-9eae4d0b.pth',
'config': './configs/htc/htc_x101_32x4d_fpn_20e_16gpu.py',
'paper_results': None}
)
model_configs.append(
{'model_name': 'HTC (ResNeXt-101 64x4d-FPN)',
'paper_arxiv_id': '1901.07518',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_x101_64x4d_fpn_20e_20190408-497f2561.pth',
'weights_name': 'htc_x101_64x4d_fpn_20e_20190408-497f2561.pth',
'config': './configs/htc/htc_x101_64x4d_fpn_20e_16gpu.py',
'paper_results': None}
)
model_configs.append(
{'model_name': 'HTC (ResNeXt-101 64x4d-FPN, DCN, multi-scale)',
'paper_arxiv_id': '1901.07518',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_dconv_c3-c5_mstrain_400_1400_x101_64x4d_fpn_20e_20190408-0e50669c.pth',
'weights_name': 'htc_dconv_c3-c5_mstrain_400_1400_x101_64x4d_fpn_20e_20190408-0e50669c.pth',
'config': './configs/htc/htc_dconv_c3-c5_mstrain_400_1400_x101_64x4d_fpn_20e.py',
'paper_results': None}
)
## Cascade Mask R-CNN
model_configs.append(
{'model_name': 'Cascade Mask R-CNN (ResNet-50-FPN, 1x LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_mask_rcnn_r50_fpn_1x_20181123-88b170c9.pth',
'weights_name': 'cascade_mask_rcnn_r50_fpn_1x_20181123-88b170c9.pth',
'config': './configs/cascade_mask_rcnn_r50_fpn_1x.py',
'paper_results': {'box AP': 0.413, 'AP50': 0.594, 'AP75': 0.453, 'APS': 0.232, 'APM': 0.438, 'APL': 0.558}}
)
model_configs.append(
{'model_name': 'Cascade Mask R-CNN (ResNet-50-FPN, 20e LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_mask_rcnn_r50_fpn_20e_20181123-6e0c9713.pth',
'weights_name': 'cascade_mask_rcnn_r50_fpn_20e_20181123-6e0c9713.pth',
'config': './configs/cascade_mask_rcnn_r50_fpn_1x.py',
'paper_results': None}
)
model_configs.append(
{'model_name': 'Cascade Mask R-CNN (ResNet-101-FPN, 1x LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_mask_rcnn_r101_fpn_1x_20181129-64f00602.pth',
'weights_name': 'cascade_mask_rcnn_r101_fpn_1x_20181129-64f00602.pth',
'config': './configs/cascade_mask_rcnn_r101_fpn_1x.py',
'paper_results': None}
)
model_configs.append(
{'model_name': 'Cascade Mask R-CNN (ResNet-101-FPN, 20e LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_mask_rcnn_r101_fpn_20e_20181129-cb85151d.pth',
'weights_name': 'cascade_mask_rcnn_r101_fpn_20e_20181129-cb85151d.pth',
'config': './configs/cascade_mask_rcnn_r101_fpn_1x.py',
'paper_results': {'box AP': 0.433, 'AP50': 0.617, 'AP75': 0.472, 'APS': 0.242, 'APM': 0.463, 'APL': 0.582}}
)
model_configs.append(
{'model_name': 'Cascade Mask R-CNN (ResNeXt-101 32x4d-FPN, 1x LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_mask_rcnn_x101_32x4d_fpn_1x_20181218-1d944c89.pth',
'weights_name': 'cascade_mask_rcnn_x101_32x4d_fpn_1x_20181218-1d944c89.pth',
'config': './configs/cascade_mask_rcnn_x101_32x4d_fpn_1x.py',
'paper_results': None}
)
model_configs.append(
{'model_name': 'Cascade Mask R-CNN (ResNeXt-101 32x4d-FPN, 20e LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_mask_rcnn_x101_32x4d_fpn_20e_20181218-761a3473.pth',
'weights_name': 'cascade_mask_rcnn_x101_32x4d_fpn_20e_20181218-761a3473.pth',
'config': './configs/cascade_mask_rcnn_x101_32x4d_fpn_1x.py',
'paper_results': None}
)
model_configs.append(
{'model_name': 'Cascade Mask R-CNN (ResNeXt-101 64x4d-FPN, 1x LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_mask_rcnn_x101_64x4d_fpn_1x_20190501-827e0a70.pth',
'weights_name': 'cascade_mask_rcnn_x101_64x4d_fpn_1x_20190501-827e0a70.pth',
'config': './configs/cascade_mask_rcnn_x101_64x4d_fpn_1x.py',
'paper_results': None}
)
model_configs.append(
{'model_name': 'Cascade Mask R-CNN (ResNeXt-101 64x4d-FPN, 20e LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_mask_rcnn_x101_64x4d_fpn_20e_20181218-630773a7.pth',
'weights_name': 'htc_dconv_c3-c5_mstrain_400_1400_x101_64x4d_fpn_20e_20190408-0e50669c.pth',
'config': './configs/cascade_mask_rcnn_x101_64x4d_fpn_1x.py',
'paper_results': {'box AP': 0.458, 'AP50': 0.641, 'AP75': 0.503, 'APS': 0.272, 'APM': 0.495, 'APL': 0.601}}
)
## Cascade R-CNN
model_configs.append(
{'model_name': 'Cascade R-CNN (ResNet-50-FPN+)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_rcnn_r50_fpn_1x_20190501-3b6211ab.pth',
'weights_name': 'cascade_rcnn_r50_fpn_1x_20190501-3b6211ab.pth',
'config': './configs/cascade_rcnn_r50_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Cascade R-CNN (ResNet-50-FPN, 20e LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_rcnn_r50_fpn_20e_20181123-db483a09.pth',
'weights_name': 'cascade_rcnn_r50_fpn_20e_20181123-db483a09.pth',
'config': './configs/cascade_rcnn_r50_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Cascade R-CNN (ResNet-101-FPN, 1x LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_rcnn_r101_fpn_1x_20181129-d64ebac7.pth',
'weights_name': 'cascade_rcnn_r101_fpn_1x_20181129-d64ebac7.pth',
'config': './configs/cascade_rcnn_r101_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Cascade R-CNN (ResNet-101-FPN+, cascade)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_rcnn_r101_fpn_20e_20181129-b46dcede.pth',
'weights_name': 'cascade_rcnn_r101_fpn_20e_20181129-b46dcede.pth',
'config': './configs/cascade_rcnn_r101_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Cascade R-CNN (ResNeXt-101 32x4d-FPN, 1x LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_rcnn_x101_32x4d_fpn_1x_20190501-af628be5.pth',
'weights_name': 'cascade_rcnn_x101_32x4d_fpn_1x_20190501-af628be5.pth',
'config': './configs/cascade_rcnn_x101_32x4d_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Cascade R-CNN (ResNeXt-101 32x4d-FPN, 20e LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_rcnn_x101_32x4d_fpn_2x_20181218-28f73c4c.pth',
'weights_name': 'cascade_rcnn_x101_32x4d_fpn_2x_20181218-28f73c4c.pth',
'config': './configs/cascade_rcnn_x101_32x4d_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Cascade R-CNN (ResNeXt-101 64x4d-FPN, 1x LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_rcnn_x101_64x4d_fpn_1x_20181218-e2dc376a.pth',
'weights_name': 'cascade_rcnn_x101_64x4d_fpn_1x_20181218-e2dc376a.pth',
'config': './configs/cascade_rcnn_x101_64x4d_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Cascade R-CNN (ResNeXt-101 64x4d-FPN, 20e LR)',
'paper_arxiv_id': '1906.09756',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/cascade_rcnn_x101_64x4d_fpn_2x_20181218-5add321e.pth',
'weights_name': 'cascade_rcnn_x101_64x4d_fpn_2x_20181218-5add321e.pth',
'config': './configs/cascade_rcnn_x101_64x4d_fpn_1x.py'}
)
## RetinaNet
model_configs.append(
{'model_name': 'RetinaNet (ResNet-50-FPN+)',
'paper_arxiv_id': '1708.02002',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/retinanet_r50_fpn_1x_20181125-7b0c2548.pth',
'weights_name': 'retinanet_r50_fpn_1x_20181125-7b0c2548.pth',
'config': './configs/retinanet_r50_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'RetinaNet (ResNet-50-FPN, 20e LR)',
'paper_arxiv_id': '1708.02002',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/retinanet_r50_fpn_2x_20190616-75574209.pth',
'weights_name': 'retinanet_r50_fpn_2x_20190616-75574209.pth',
'config': './configs/retinanet_r50_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'RetinaNet (ResNet-101-FPN, 1x LR)',
'paper_arxiv_id': '1708.02002',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/retinanet_r101_fpn_1x_20181129-f016f384.pth',
'weights_name': 'retinanet_r101_fpn_1x_20181129-f016f384.pth',
'config': './configs/retinanet_r101_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'RetinaNet (ResNet-101-FPN+, cascade)',
'paper_arxiv_id': '1708.02002',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/retinanet_r101_fpn_2x_20181129-72c14526.pth',
'weights_name': 'retinanet_r101_fpn_2x_20181129-72c14526.pth',
'config': './configs/retinanet_r101_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'RetinaNet (ResNeXt-101 32x4d-FPN, 1x LR)',
'paper_arxiv_id': '1708.02002',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/retinanet_x101_32x4d_fpn_1x_20190501-967812ba.pth',
'weights_name': 'retinanet_x101_32x4d_fpn_1x_20190501-967812ba.pth',
'config': './configs/retinanet_x101_32x4d_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'RetinaNet (ResNeXt-101 32x4d-FPN, 20e LR)',
'paper_arxiv_id': '1708.02002',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/retinanet_x101_32x4d_fpn_2x_20181218-8596452d.pth',
'weights_name': 'retinanet_x101_32x4d_fpn_2x_20181218-8596452d.pth',
'config': './configs/retinanet_x101_32x4d_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'RetinaNet (ResNeXt-101 64x4d-FPN, 1x LR)',
'paper_arxiv_id': '1708.02002',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/retinanet_x101_64x4d_fpn_1x_20181218-a0a22662.pth',
'weights_name': 'retinanet_x101_64x4d_fpn_1x_20181218-a0a22662.pth',
'config': './configs/retinanet_x101_64x4d_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'RetinaNet (ResNeXt-101 64x4d-FPN, 20e LR)',
'paper_arxiv_id': '1708.02002',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/retinanet_x101_64x4d_fpn_2x_20181218-5e88d045.pth',
'weights_name': 'retinanet_x101_64x4d_fpn_2x_20181218-5e88d045.pth',
'config': './configs/retinanet_x101_64x4d_fpn_1x.py'}
)
## Faster R-CNN
model_configs.append(
{'model_name': 'Faster R-CNN (ResNet-50-FPN, 1x LR)',
'paper_arxiv_id': '1506.01497',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/faster_rcnn_r50_fpn_1x_20181010-3d1b3351.pth',
'weights_name': 'faster_rcnn_r50_fpn_1x_20181010-3d1b3351.pth',
'config': './configs/faster_rcnn_r50_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Faster R-CNN (ResNet-50-FPN, 2x LR)',
'paper_arxiv_id': '1506.01497',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/faster_rcnn_r50_fpn_2x_20181010-443129e1.pth',
'weights_name': 'faster_rcnn_r50_fpn_2x_20181010-443129e1.pth',
'config': './configs/faster_rcnn_r50_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Faster R-CNN (ResNet-101-FPN, 1x LR)',
'paper_arxiv_id': '1506.01497',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/faster_rcnn_r101_fpn_1x_20181129-d1468807.pth',
'weights_name': 'faster_rcnn_r101_fpn_1x_20181129-d1468807.pth',
'config': './configs/faster_rcnn_r101_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Faster R-CNN (ResNet-101-FPN, 2x LR)',
'paper_arxiv_id': '1506.01497',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/faster_rcnn_r101_fpn_2x_20181129-73e7ade7.pth',
'weights_name': 'faster_rcnn_r101_fpn_2x_20181129-73e7ade7.pth',
'config': './configs/faster_rcnn_r101_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Faster R-CNN (ResNeXt-101 32x4d-FPN, 1x LR)',
'paper_arxiv_id': '1506.01497',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/faster_rcnn_x101_32x4d_fpn_1x_20181218-ad81c133.pth',
'weights_name': 'faster_rcnn_x101_32x4d_fpn_1x_20181218-ad81c133.pth',
'config': './configs/faster_rcnn_x101_32x4d_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Faster R-CNN (ResNeXt-101 32x4d-FPN, 2x LR)',
'paper_arxiv_id': '1506.01497',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/faster_rcnn_x101_32x4d_fpn_2x_20181218-0ed58946.pth',
'weights_name': 'faster_rcnn_x101_32x4d_fpn_2x_20181218-0ed58946.pth',
'config': './configs/faster_rcnn_x101_32x4d_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Faster R-CNN (ResNeXt-101 64x4d-FPN, 1x LR)',
'paper_arxiv_id': '1506.01497',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/faster_rcnn_x101_64x4d_fpn_1x_20181218-c9c69c8f.pth',
'weights_name': 'faster_rcnn_x101_64x4d_fpn_1x_20181218-c9c69c8f.pth',
'config': './configs/faster_rcnn_x101_64x4d_fpn_1x.py'}
)
model_configs.append(
{'model_name': 'Faster R-CNN (ResNeXt-101 64x4d-FPN, 2x LR)',
'paper_arxiv_id': '1506.01497',
'weights_url': 'https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/faster_rcnn_x101_64x4d_fpn_2x_20181218-fe94f9b8.pth',
'weights_name': 'faster_rcnn_x101_64x4d_fpn_2x_20181218-fe94f9b8.pth',
'config': './configs/faster_rcnn_x101_64x4d_fpn_1x.py'}
)
## HRNet
model_configs.append(
{'model_name': 'Faster R-CNN (HRNetV2p-W18, 1x LR)',
'paper_arxiv_id': '1904.04514',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/hrnet/faster_rcnn_hrnetv2p_w18_1x_20190522-e368c387.pth',
'weights_name': 'faster_rcnn_hrnetv2p_w18_1x_20190522-e368c387.pth',
'config': './configs/hrnet/faster_rcnn_hrnetv2p_w18_1x.py',
'paper_results': {'box AP': 0.362, 'AP50': 0.573, 'AP75': 0.393, 'APS': 0.207, 'APM': 0.39, 'APL': 0.468}}
)
model_configs.append(
{'model_name': 'Faster R-CNN (HRNetV2p-W18)',
'paper_arxiv_id': '1904.04514',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/hrnet/faster_rcnn_hrnetv2p_w18_2x_20190810-9c8615d5.pth',
'weights_name': 'faster_rcnn_hrnetv2p_w18_2x_20190810-9c8615d5.pth',
'config': './configs/hrnet/faster_rcnn_hrnetv2p_w18_1x.py'}
)
model_configs.append(
{'model_name': 'Faster R-CNN (HRNetV2p-W32, 1x LR)',
'paper_arxiv_id': '1904.04514',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/hrnet/faster_rcnn_hrnetv2p_w32_1x_20190522-d22f1fef.pth',
'weights_name': 'faster_rcnn_hrnetv2p_w32_1x_20190522-d22f1fef.pth',
'config': './configs/hrnet/faster_rcnn_hrnetv2p_w32_1x.py',
'paper_results': {'box AP': 0.396, 'AP50': 0.61, 'AP75': 0.433, 'APS': 0.237, 'APM': 0.425, 'APL': 0.505}}
)
model_configs.append(
{'model_name': 'Faster R-CNN (HRNetV2p-W32)',
'paper_arxiv_id': '1904.04514',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/hrnet/faster_rcnn_hrnetv2p_w32_2x_20190810-24e8912a.pth',
'weights_name': 'faster_rcnn_hrnetv2p_w32_2x_20190810-24e8912a.pth',
'config': './configs/hrnet/faster_rcnn_hrnetv2p_w32_1x.py'}
)
#maskrcnn
model_configs.append(
{'model_name': 'Mask R-CNN (HRNetV2p-W18, 1x LR)',
'paper_arxiv_id': '1904.04514',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/hrnet/mask_rcnn_hrnetv2p_w18_1x_20190522-c8ad459f.pth',
'weights_name': 'mask_rcnn_hrnetv2p_w18_1x_20190522-c8ad459f.pth',
'config': './configs/hrnet/mask_rcnn_hrnetv2p_w18_1x.py',
'paper_results': {'box AP': 0.371, 'APS': 0.219, 'APM': 0.395, 'APL': 0.479}}
)
model_configs.append(
{'model_name': 'Mask R-CNN (HRNetV2p-W18)',
'paper_arxiv_id': '1904.04514',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/hrnet/mask_rcnn_hrnetv2p_w18_2x_20190810-1e4747eb.pth',
'weights_name': 'mask_rcnn_hrnetv2p_w18_2x_20190810-1e4747eb.pth',
'config': './configs/hrnet/mask_rcnn_hrnetv2p_w18_1x.py'}
)
model_configs.append(
{'model_name': 'Mask R-CNN (HRNetV2p-W32, 1x LR)',
'paper_arxiv_id': '1904.04514',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/hrnet/mask_rcnn_hrnetv2p_w32_1x_20190522-374aaa00.pth',
'weights_name': 'mask_rcnn_hrnetv2p_w32_1x_20190522-374aaa00.pth',
'config': './configs/hrnet/mask_rcnn_hrnetv2p_w32_1x.py',
'paper_results': {'box AP': 0.409, 'APS': 0.245, 'APM': 0.439, 'APL': 0.522}}
)
model_configs.append(
{'model_name': 'Mask R-CNN (HRNetV2p-W32)',
'paper_arxiv_id': '1904.04514',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/hrnet/mask_rcnn_hrnetv2p_w32_2x_20190810-773eca75.pth',
'weights_name': 'mask_rcnn_hrnetv2p_w32_2x_20190810-773eca75.pth',
'config': './configs/hrnet/mask_rcnn_hrnetv2p_w32_1x.py'}
)
#cascadercnn
model_configs.append(
{'model_name': 'Cascade R-CNN (HRNetV2p-W32)',
'paper_arxiv_id': '1904.04514',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/hrnet/cascade_rcnn_hrnetv2p_w32_20e_20190522-55bec4ee.pth',
'weights_name': 'cascade_rcnn_hrnetv2p_w32_20e_20190522-55bec4ee.pth',
'config': './configs/hrnet/cascade_rcnn_hrnetv2p_w32_20e.py'}
)
#cascademaskkrcnn
model_configs.append(
{'model_name': 'Mask R-CNN (HRNetV2p-W32, cascade)',
'paper_arxiv_id': '1904.04514',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/hrnet/cascade_mask_rcnn_hrnetv2p_w32_20e_20190810-76f61cd0.pth',
'weights_name': 'cascade_mask_rcnn_hrnetv2p_w32_20e_20190810-76f61cd0.pth',
'config': './configs/hrnet/cascade_mask_rcnn_hrnetv2p_w32_20e.py'}
)
#htc
model_configs.append(
{'model_name': 'HTC (HRNetV2p-W32)',
'paper_arxiv_id': '1904.04514',
'weights_url': 'https://open-mmlab.s3.ap-northeast-2.amazonaws.com/mmdetection/models/hrnet/htc_hrnetv2p_w32_20e_20190810-82f9ef5a.pth',
'weights_name': 'htc_hrnetv2p_w32_20e_20190810-82f9ef5a.pth',
'config': './configs/hrnet/htc_hrnetv2p_w32_20e.py'}
)
import torch.distributed as dist
dist.init_process_group('gloo', init_method='file:///tmp/somefile', rank=0, world_size=1)
for model_config in model_configs:
if 'paper_results' not in model_config:
model_config['paper_results'] = None
evaluate_model(model_name=model_config['model_name'],
paper_arxiv_id=model_config['paper_arxiv_id'],
weights_url=model_config['weights_url'],
weights_name=model_config['weights_name'],
paper_results=model_config['paper_results'],
config=model_config['config'])
STATUS
BUILD
COMMIT MESSAGE
RUN TIME