search

microsoft / scene_graph_benchmark

image scene graph generation benchmark
MIT License
382 stars 86 forks source link

Guide to run train_net.py #28

Open Edwardmark opened 3 years ago

Edwardmark commented 3 years ago

Hi, I want to train the detector with train_net.py, could you please give me some guide? How to organize the data, and how to pass the parameters. Thanks. @pzzhang

Edwardmark commented 3 years ago

@hanxiaotian For example, if I want to fine-tune the model with only coco dataset, what command should be run? And what modification should be made to the code? Could you please give some detailed guidance? Thanks in advance.

Edwardmark commented 3 years ago

I follow the tsv dataset demo to generate coco dataset in tsv format, but when I run the train_net.py, https://github.com/microsoft/scene_graph_benchmark/blob/a93180e85ce69baf67cd6300731baff8d10c1160/maskrcnn_benchmark/data/datasets/od_tsv.py#L46

the above line contains annotations['objects'], but the code: https://github.com/microsoft/scene_graph_benchmark/blob/a93180e85ce69baf67cd6300731baff8d10c1160/tools/mini_tsv/tsv_demo.py#L37

there is no key named "objects". Any clues? Thanks. @hanxiaotian @pzzhang