Add support for image absolute path by using the "parent" directory in the data.yaml #2071
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
added 6 commits
August 7, 2024 18:03
Author
|
check train dataset objects class id violation. n-ch in as param. |
added 23 commits
September 15, 2024 09:20
…B , inversion aug is dtype dependant. test.py : verbose save class mAP , mAP per size. ClearML support
gradient clipping with optimizer scaler CLearML connect config
remove no_aug
scaling before mosaic with predefined filling_value according to img precision chnaged by scaling before mosaic. filling uint16 with image.mean() detect.py : align to latest changes as test.py test.py : support in large object size metrices stats_all_large plots.py : plot detections with th>0.1 (not 0.25)
…uce edges artifacts. mAP per med/large exclusively
added 30 commits
January 6, 2025 17:28
… issue the prediction over the random padding area should be omitted from calculation
…ith metadata of each image, caching labels omit label w/ class no not in the list. range_bar plot, add plot class support. Onnx RT: adding noise to dataset, dataset with ext. noise
…ange p/r stats, height based range, p/r per range per class
…right crop. Whether condition based breakdown
…ds but GT exist. whether statistics, ONNX save pre-processed file,
FL with configured weighted loss : alpha test.py: tSNE over the cropped bbox features
FL with configured weighted loss : alpha test.py: tSNE over the cropped bbox features
… fix effective height computation datasets.py : Remove empty annotated image entries weights file moved to HDD
Trigger when AP fall
…nter_roi_no_swiss_no_38B_20b_22c_n_png_train_3_cls_jan_25.txt
training-set Test36B 12/11 +missing png/s and other tiff from the final annotation list : tir_tiff_seq_png_3_class_fixed_whether_copied_dataset_label.xlsx
training-set Test36B 12/11 +missing png/s and other tiff from the final annotation list : tir_tiff_seq_png_3_class_fixed_whether_copied_dataset_label.xlsx
+missing png/s and other tiff from the final annotation list : tir_tiff_seq_png_3_class_fixed_whether_copied_dataset_label.xlsx
…hted loss Adding Asymentrical loss for Multi-label hasn;t tested yet!
Fix : random seed test.py : accumulate loss per image hard to run natchwise
best_model weights most the mAP@50 summary reposrter: merge_results.py of best model remove aggresive truncated objects center crop from training list
Tensor-leap integration update requirements test.py : fix undetected images handling in breakdown analysis
integration with Tleap
asymetric loss has bug yielding nan Created unlabeld seq file extractor
plots.py - add image plot to clearML
plots.py - add image plot to clearML
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In the 2nd option : file: path/images.txt the "path: " can serve as an absolute path for the image folders given inside "train :" field rather than the current option strict which is the parent folder.