search

Eli-YiLi / WSSS_MMSeg

Pseudo-mask Matters in Weakly-supervised Semantic Segmentation
Apache License 2.0
18 stars 3 forks source link

Question about COCO Pseudo Mask #6

Closed Unrealluver closed 2 years ago

Unrealluver commented 2 years ago

Hi there! Your work is so awesome at the performence! I'd like to do something on it. But when I run the code you commited, I meet some question:

  1. What's the label '0' refer to in your ppmg_coco pseudo mask? I found that people and the background are all labeled as '0' in some pseudo masks like 000000000036.png.
  2. Is that your label mapping is showed as bellow?

CLASSES = ( 'person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus', 'train', 'truck', 'boat', 'traffic light', 'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog', 'horse', 'sheep', 'cow', 'elephant', 'bear', 'zebra', 'giraffe', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase', 'frisbee', 'skis', 'snowboard', 'sports ball', 'kite', 'baseball bat', 'baseball glove', 'skateboard', 'surfboard', 'tennis racket', 'bottle', 'wine glass', 'cup', 'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple', 'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza', 'donut', 'cake', 'chair', 'couch', 'potted plant', 'bed', 'dining table', 'toilet', 'tv', 'laptop', 'mouse', 'remote', 'keyboard', 'cell phone', 'microwave', 'oven', 'toaster', 'sink', 'refrigerator', 'book', 'clock', 'vase', 'scissors', 'teddy bear', 'hair drier', 'toothbrush', 'banner', 'blanket', 'branch', 'bridge', 'building-other', 'bush', 'cabinet', 'cage', 'cardboard', 'carpet', 'ceiling-other', 'ceiling-tile', 'cloth', 'clothes', 'clouds', 'counter', 'cupboard', 'curtain', 'desk-stuff', 'dirt', 'door-stuff', 'fence', 'floor-marble', 'floor-other', 'floor-stone', 'floor-tile', 'floor-wood', 'flower', 'fog', 'food-other', 'fruit', 'furniture-other', 'grass', 'gravel', 'ground-other', 'hill', 'house', 'leaves', 'light', 'mat', 'metal', 'mirror-stuff', 'moss', 'mountain', 'mud', 'napkin', 'net', 'paper', 'pavement', 'pillow', 'plant-other', 'plastic', 'platform', 'playingfield', 'railing', 'railroad', 'river', 'road', 'rock', 'roof', 'rug', 'salad', 'sand', 'sea', 'shelf', 'sky-other', 'skyscraper', 'snow', 'solid-other', 'stairs', 'stone', 'straw', 'structural-other', 'table', 'tent', 'textile-other', 'towel', 'tree', 'vegetable', 'wall-brick', 'wall-concrete', 'wall-other', 'wall-panel', 'wall-stone', 'wall-tile', 'wall-wood', 'water-other', 'waterdrops', 'window-blind', 'window-other', 'wood')

Unrealluver commented 2 years ago

looking for your reply, thanks

Unrealluver commented 2 years ago

Do you use some chatting apps like wechat? Could we talk about more detailed things on it?

Eli-YiLi commented 2 years ago
  1. 0 is background.
  2. label_map: https://tech.amikelive.com/node-718/what-object-categories-labels-are-in-coco-dataset/ (range from 0-91 with bg, only 81 classes are valid)
Unrealluver commented 2 years ago

thanks!

Unrealluver commented 2 years ago

But I still have question that why the pseudo mask in foider ppmg_coco mislead the class background and people in a very simple image 000000000036? And I also checked other images that contain people, their psudo mask also labeled people area as '0'.

Unrealluver commented 2 years ago

Thanks for the blog that truly helped me unstand the coco14 dataset! I noticed that the invalid classes' num in the blog is 11 and the valid classes contains 80 ones. So should we set the coco14's class_num to 92 (80 valid foregrounds + 11 invalid classes + 1 background) to get the best CAM performence?

Unrealluver commented 2 years ago

Thanks for the blog that truly helped me unstand the coco14 dataset! I noticed that the invalid classes' num in the blog is 11 and the valid classes contains 80 ones. So should we set the coco14's class_num to 92 (80 valid foregrounds + 11 invalid classes + 1 background) to get the best CAM performence?

Sorry, I noticed that the last invalid is not need to be add. QAQ