search

NeurAI-Lab / MMC

The official PyTorch code for NeurIPS 2021 ML4AD Paper, "Does Thermal data make the detection systems more reliable?"
MIT License
13 stars 2 forks source link

About data set label format #2

Closed wlc-git closed 2 years ago

wlc-git commented 2 years ago

'KAIST_COMBINed_COCO_train_set611 ': { "Data_dir" : "" train"," "Ann_file" : "train/sanitized_annotations/annotations/instances_comb json." " }, Which part of data is it? Is your JSON encapsulated XML or YOLO format? Are the train and test parts of KAIST data set used cleaned data Looking forward to your reply

shruum commented 2 years ago

Hi, thanks for your response.

The JSON is in XML COCO format. The original annotations included few errors and were also redundant, and hence we used a more sanitized set defines by Multispectral Pedestrian Detection via Simultaneous Detection and Segmentation