search

XinGP / DGFNet

Multi-Agent Trajectory Prediction with Difficulty-Guided Feature Enhancement Network
Apache License 2.0
37 stars 5 forks source link

Test for leaderboard submission #2

Open alien-HL opened 2 months ago

alien-HL commented 2 months ago

Nice code , thanks for your work , I need a test code for leaderboard submission , can you provide some examples please?

XinGP commented 2 months ago

We used Lanegcn's test code before, and the code address is https://github.com/uber-research/LaneGCN

John122-art commented 18 hours ago

thanks for your work,the result of test.py is a documents with the attrbutes of h5,but we need to submit json document on EvalAI,how do you slove this problem?

XinGP commented 17 hours ago

Sorry, I didn't quite understand your question. EvalAI can upload files in H5 format normally. Why do you need to convert them to JSON format?

John122-art commented 17 hours ago

Sorry, I didn't quite understand your question. EvalAI can upload files in H5 format normally. Why do you need to convert them to JSON format?

image image image I followed the steps above and submitted the corresponding .h5 file, but an error occurred. So, I asked ChatGPT, and it replied that the file type was incorrect.

John122-art commented 16 hours ago

I chose the wrong competition category, and I apologize for any inconvenience this may have caused.

John122-art commented 5 hours ago

I'm sorry to inform you that I trained and saved the final results using your code, and after processing the .h5 file with test.py and submitting it to EvalAI, I found a large discrepancy in the results. My initial suspicion is that the argo_preprocessed file underwent two translations and rotations, while in test.py, only one rotation was used for the restoration. Perhaps my assumption is incorrect. Could you please explain this to me? Thank you very much. image image