@@ -387,53 +387,176 @@ for img, labels in loader:
387387The supported formats are:
388388
389389- ** COCO** - We support COCO JSON format in two variants:
390+
391+ - [ FiftyOne layout] ( https://docs.voxel51.com/user_guide/export_datasets.html#cocodetectiondataset-export )
392+ ``` plaintext
393+ dataset_dir/
394+ ├── train/
395+ │ ├── data/
396+ │ │ ├── img1.jpg
397+ │ │ ├── img2.jpg
398+ │ │ └── ...
399+ │ └── labels.json
400+ ├── validation/
401+ │ ├── data/
402+ │ └── labels.json
403+ └── test/
404+ ├── data/
405+ └── labels.json
406+ ```
390407 - [RoboFlow](https://roboflow.com/formats/coco-json)
391- - [ FiftyOne] ( https://docs.voxel51.com/user_guide/export_datasets.html#cocodetectiondataset-export )
408+ ```plaintext
409+ dataset_dir/
410+ ├── train/
411+ │ ├── img1.jpg
412+ │ ├── img2.jpg
413+ │ └── ...
414+ │ └── _annotations.coco.json
415+ ├── valid/
416+ └── test/
417+ ```
418+
392419- [**Pascal VOC XML**](https://roboflow.com/formats/pascal-voc-xml)
420+
421+ ```plaintext
422+ dataset_dir/
423+ ├── train/
424+ │ ├── img1.jpg
425+ │ ├── img1.xml
426+ │ └── ...
427+ ├── valid/
428+ └── test/
429+ ```
430+
393431- [ ** YOLO Darknet TXT** ] ( https://roboflow.com/formats/yolo-darknet-txt )
432+
433+ ``` plaintext
434+ dataset_dir/
435+ ├── train/
436+ │ ├── img1.jpg
437+ │ ├── img1.txt
438+ │ ├── ...
439+ │ └── _darknet.labels
440+ ├── valid/
441+ └── test/
442+ ```
443+
394444- [ ** YOLOv4 PyTorch TXT** ] ( https://roboflow.com/formats/yolov4-pytorch-txt )
445+
446+ ``` plaintext
447+ dataset_dir/
448+ ├── train/
449+ │ ├── img1.jpg
450+ │ ├── img2.jpg
451+ │ ├── ...
452+ │ ├── _annotations.txt
453+ │ └── _classes.txt
454+ ├── valid/
455+ └── test/
456+ ```
457+
395458- [ ** MT YOLOv6** ] ( https://roboflow.com/formats/mt-yolov6 )
459+
460+ ``` plaintext
461+ dataset_dir/
462+ ├── images/
463+ │ ├── train/
464+ │ │ ├── img1.jpg
465+ │ │ ├── img2.jpg
466+ │ │ └── ...
467+ │ ├── valid/
468+ │ └── test/
469+ ├── labels/
470+ │ ├── train/
471+ │ │ ├── img1.txt
472+ │ │ ├── img2.txt
473+ │ │ └── ...
474+ │ ├── valid/
475+ │ └── test/
476+ └── data.yaml
477+ ```
478+
396479- [ ** CreateML JSON** ] ( https://roboflow.com/formats/createml-json )
480+
481+ ``` plaintext
482+ dataset_dir/
483+ ├── train/
484+ │ ├── img1.jpg
485+ │ ├── img2.jpg
486+ │ └── ...
487+ │ └── _annotations.createml.json
488+ ├── valid/
489+ └── test/
490+ ```
491+
397492- [ ** TensorFlow Object Detection CSV** ] ( https://roboflow.com/formats/tensorflow-object-detection-csv )
493+
494+ ``` plaintext
495+ dataset_dir/
496+ ├── train/
497+ │ ├── img1.jpg
498+ │ ├── img2.jpg
499+ │ ├── ...
500+ │ └── _annotations.csv
501+ ├── valid/
502+ └── test/
503+ ```
504+
398505- [ ** SOLO
** ] ( https://docs.unity3d.com/Packages/[email protected] /manual/Schema/SoloSchema.html ) 399- - ** Classification Directory** - A directory with subdirectories for each class
400506
401- ``` plaintext
402- dataset_dir/
403- ├── train/
404- │ ├── class1/
405- │ │ ├── img1.jpg
406- │ │ ├── img2.jpg
407- │ │ └── ...
408- │ ├── class2/
409- │ └── ...
410- ├── valid/
411- └── test/
412- ```
507+ ``` plaintext
508+ dataset_dir/
509+ ├── train/
510+ │ ├── metadata.json
511+ │ ├── sensor_definitions.json
512+ │ ├── annotation_definitions.json
513+ │ ├── metric_definitions.json
514+ │ └── sequence.<SequenceNUM>/
515+ │ ├── step<StepNUM>.camera.jpg
516+ │ ├── step<StepNUM>.frame_data.json
517+ │ └── (OPTIONAL: step<StepNUM>.camera.semantic segmentation.jpg)
518+ ├── valid/
519+ └── test/
520+ ```
413521
414- - ** Segmentation Mask Directory** - A directory with images and corresponding masks.
522+ - ** Classification Directory** - A directory with subdirectories for each class
415523
416- ``` plaintext
417- dataset_dir/
418- ├── train/
419- │ ├── img1.jpg
420- │ ├── img1_mask.png
421- │ ├── ...
422- │ └── _classes.csv
423- ├── valid/
424- └── test/
425- ```
524+ ``` plaintext
525+ dataset_dir/
526+ ├── train/
527+ │ ├── class1/
528+ │ │ ├── img1.jpg
529+ │ │ ├── img2.jpg
530+ │ │ └── ...
531+ │ ├── class2/
532+ │ └── ...
533+ ├── valid/
534+ └── test/
535+ ```
426536
427- The masks are stored as grayscale PNG images where each pixel value corresponds to a class.
428- The mapping from pixel values to class is defined in the ` _classes.csv ` file.
537+ - ** Segmentation Mask Directory** - A directory with images and corresponding masks.
429538
430- ``` csv
431- Pixel Value, Class
432- 0, background
433- 1, class1
434- 2, class2
435- 3, class3
436- ```
539+ ``` plaintext
540+ dataset_dir/
541+ ├── train/
542+ │ ├── img1.jpg
543+ │ ├── img1_mask.png
544+ │ ├── ...
545+ │ └── _classes.csv
546+ ├── valid/
547+ └── test/
548+ ```
549+
550+ The masks are stored as grayscale PNG images where each pixel value corresponds to a class.
551+ The mapping from pixel values to class is defined in the ` _classes.csv ` file.
552+
553+ ``` csv
554+ Pixel Value, Class
555+ 0, background
556+ 1, class1
557+ 2, class2
558+ 3, class3
559+ ```
437560
438561### Dataset Creation
439562
0 commit comments