search

dbash / zerowaste

[CVPR 2022] ZeroWaste Dataset: Towards Deformable Object Segmentation in Cluttered Scenes
MIT License
42 stars 9 forks source link

Problems with visualisation of Deeplab evaluation results #11

Open saurmart opened 1 year ago

saurmart commented 1 year ago

Hi @dbash

after running the evaluation of the pre-trained deeplab ZeroWaste using the following code:

!python /content/drive/MyDrive/Evaluation_ZeroWaste_Deeplab/train_net.py --config-file /content/drive/MyDrive/Evaluation_ZeroWaste_Deeplab/zerowaste_config. yaml --dataroot /content/drive/MyDrive/Evaluation_ZeroWaste_Deeplab/ZeroWasteAug --eval-only OUTPUT_DIR /content/drive/MyDrive/Evaluation_ZeroWaste_Deeplab/outputs/results_mod/ MODEL.WEIGHTS /content/drive/MyDrive/Evaluation_ZeroWaste_Deeplab/model_final_aug.pth

I have the sem_seg_predictions.json which contains the segmentations for each image of the ZeroWasteAug test dataset. Here is an extract from the file:

[{
        "file_name": "/content/drive/MyDrive/Evaluation_ZeroWaste_Deeplab/ZeroWasteAug/test/data/01_frame_000680.PNG",
        "category_id": 0,
        "segmentation": {
            "size": [1080, 1920],
            "counts": "0ja0T2j[...]0VYO"
        }
    },

Now I want to visualise the results, like plotting the predictions of some images to compare them with the GT images. But I am not sure how to do this because I do not know how to deal with this sem_seg_predictions.json? Do you have any code to do this task. I would really appreciate it.