2c3efa430b
Mosaic plots bug fix ( #1526 )
4 years ago
12499f1c01
--image_weights bug fix ( #1524 )
4 years ago
9728e2b8ae
--image_weights bug fix ( #1524 )
4 years ago
e9a0ae6f19
Cache bug fix ( #1513 )
* Caching bug fix #1508
* np.zeros((0,5)) x2
4 years ago
b3ceffb513
Add QFocalLoss() ( #1482 )
* Update loss.py
implement the quality focal loss which is a more general case of focal loss
more detail in https://arxiv.org/abs/2006.04388
In the obj loss (or the case cls loss with label smooth), the targets is no long barely be 0 or 1 (can be 0.7), in this case, the normal focal loss is not work accurately
quality focal loss in behave the same as focal loss when the target is equal to 0 or 1, and work accurately when targets in (0, 1)
example:
targets:
tensor([[0.6225, 0.0000, 0.0000],
[0.9000, 0.0000, 0.0000],
[1.0000, 0.0000, 0.0000]])
___________________________
pred_prob:
tensor([[0.6225, 0.2689, 0.1192],
[0.7773, 0.5000, 0.2227],
[0.8176, 0.8808, 0.1978]])
____________________________
focal_loss
tensor([[0.0937, 0.0328, 0.0039],
[0.0166, 0.1838, 0.0199],
[0.0039, 1.3186, 0.0145]])
______________
qfocal_loss
tensor([[7.5373e-08, 3.2768e-02, 3.9179e-03],
[4.8601e-03, 1.8380e-01, 1.9857e-02],
[3.9233e-03, 1.3186e+00, 1.4545e-02]])
we can see that targets[0][0] = 0.6255 is almost the same as pred_prob[0][0] = 0.6225,
the targets[1][0] = 0.9 is greater then pred_prob[1][0] = 0.7773 by 0.1227
however, the focal loss[0][0] = 0.0937 larger then focal loss[1][0] = 0.0166 (which against the purpose of focal loss)
for the quality focal loss , it implement the case of targets not equal to 0 or 1
* Update loss.py
4 years ago
2026d4c5eb
Update caching ( #1496 )
4 years ago
bde5d9aaaa
Update caching ( #1496 )
4 years ago
0822cda781
Update caching ( #1496 )
4 years ago
89c7a5b8dc
Update caching ( #1496 )
4 years ago
44f42b1589
changed prints to logging in utils/datasets ( #1315 )
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
4 years ago
7aeef2dca5
Prevent PR plotting ( #1489 )
4 years ago
354109c54c
Autosplit ( #1488 )
4 years ago
4798e66fdf
Autosplit ( #1488 )
4 years ago
0a3ff71ae0
Confusion matrix ( #1474 )
* initial commit
* add plotting
* matrix to cpu
* bug fix
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* seaborn pandas to requirements.txt
* seaborn pandas to requirements.txt
* update wandb plotting
* remove pandas
* if plots
* if plots
* if plots
* if plots
* if plots
* initial commit
* add plotting
* matrix to cpu
* bug fix
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* update plot
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* seaborn pandas to requirements.txt
* seaborn pandas to requirements.txt
* update wandb plotting
* remove pandas
* if plots
* if plots
* if plots
* if plots
* if plots
* Cat apriori to autolabels
* cleanup
4 years ago
95fa65339f
Cat apriori to autolabels ( #1484 )
4 years ago
201bafc7cf
Sync train and test iou_thresh ( #1465 )
* Sync train and test iou_thresh
* Sync train and test iou_thresh
* weights names .lower()
* Notebook update
4 years ago
a70e554fc3
Remove redundant downloads mirror ( #1461 )
4 years ago
394131c2aa
Use torchvision.ops.nms ( #1460 )
4 years ago
05a955a3f6
FLOPS computation device bug fix ( #1447 )
* Update torch_utils.py
fix issue#113 , inputs device should be same with model parameters' device
* Update torch_utils.py
* Update torch_utils.py
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
4 years ago
64bce3c822
Update labels.png with rectangles fix ( #1432 )
4 years ago
1542ccadb2
Update labels.png with rectangles ( #1432 )
4 years ago
4250f84dfb
Update PR curve ( #1428 )
* Update PR curve
* legend outside
* list(Path().glob())
4 years ago
8d2d6d2349
Update img2label_paths() ( #1414 )
4 years ago
f5429260ca
PyTorch Hub and autoShape update ( #1415 )
* PyTorch Hub and autoShape update
* comment x for imgs
* reduce comment
4 years ago
92c9b72832
Matplotlib AGG backend ( #1413 )
4 years ago
548a98a391
Recursive directories dataset capability ( #1408 )
* Recursive directories dataset capability
* x.split('.')[-1]
* f += glob.glob(f"{p}{os.sep}**.*", recursive=True)
* f += glob.glob(str(p / '**.*'), recursive=True)
* remove . from image and vid formats
* .txt to txt
* str(p / '**' / '*.*')
4 years ago
9c91aeae10
W&B logging add hyperparameters ( #1399 )
* W&B logging add hyperparameters
* hyp bug fix and image logging updates
* if plots and wandb:
* cleanup
* wandb/ gitignore add
* cleanup 2
* cleanup 3
* move wandb import to top of file
* wandb evolve
* update import
* wandb.run.finish()
* default anchors: 3
4 years ago
9f27902368
Improved FLOPS computation ( #1398 )
4 years ago
d8f5fcfe87
Improved FLOPS computation ( #1398 )
* Improved FLOPS computation
* update comment
4 years ago
fe341fa44d
Utils reorganization ( #1392 )
* Utils reorganization
* Add new utils files
* cleanup
* simplify
* reduce datasets.py
* remove evolve.sh
* loadWebcam cleanup
4 years ago
c4addd7761
Unified '/project/name' results saving ( #1377 )
* Project/name update
* Update ci-testing.yml
* address project with path separator failure mode
* Project/name update
* address project with path separator failure mode
* Update ci-testing.yml
* detect.py default --name bug fix
* missing rstrip PR
* train/exp0 to train/exp
4 years ago
67bf9a974e
Replace 'ground truth' with 'labels' ( #1337 )
* Replace 'ground truth' with 'labels'
* Using torch... update
4 years ago
19e2482458
PyTorch version to screen and cleanup ( #1325 )
* Create flatten_recursive() helper function
* cleanup
* print torch version
4 years ago
4821d076e2
Increment train, test, detect runs/ ( #1322 )
* Increment train, test, detect runs/
* Update ci-testing.yml
* inference/images to data/images
* move images
* runs/exp to runs/train/exp
* update 'results saved to %s' str
4 years ago
d3e7778151
imap bug fix
4 years ago
194f16844e
Multi-threaded image caching
4 years ago
ea7e78cb11
Grid indices overflow bug fix ( #1316 )
4 years ago
fed5d9cdfb
Grid indices overflow bug fix ( #1313 )
4 years ago
2062765e06
auto-download v3.1 models ( #1259 )
4 years ago
c8c5ef36c9
PyTorch 1.7.0 Compatibility Updates ( #1233 )
* torch 1.7.0 compatibility updates
* add inference verification
4 years ago
453acdec67
Update tensorboard logging
4 years ago
8e9b3d0205
Precision-Recall Curve feature update #1206
4 years ago
ed85038c38
Precision-Recall Curve feature update ( #1206 )
* Precision-Recall Curve feature update
* sentinel value update
4 years ago
481d46cffb
Improved corruption handling during scan and cache ( #999 )
4 years ago
d61930e017
Improved corruption handling during scan and cache ( #999 )
4 years ago
9291daa6f7
fix padding for rectangular inference ( #1165 )
Co-authored-by: swsong <swsong@stratosphere.mobi>
4 years ago
3b57cb5641
Simplified inference ( #1153 )
4 years ago
c67e72200e
fix compatibility for hyper config ( #1146 )
* fix/hyper
* Hyp giou check to train.py
* restore general.py
* train.py overwrite fix
* restore general.py and pep8 update
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
4 years ago
4d3680c81d
Minor import and spelling updates ( #1133 )
4 years ago
00917a6225
update expt name comment and folder parsing for training ( #978 )
* comment
* fix parsing
* fix evolve
* folder
* tqdm
* Update train.py
* Update train.py
* reinstate anchors into meta dict
anchor evolution is working correctly now
* reinstate logger
prefer the single line readout for concise logging, which helps simplify notebook and tutorials etc.
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
4 years ago
Glenn Jocher
yxNONG
igornishka
yujun
Jacobsolawetz
Shiwei Song
Jirka Borovec