search

opendilab / InterFuser

[CoRL 2022] InterFuser: Safety-Enhanced Autonomous Driving Using Interpretable Sensor Fusion Transformer
Apache License 2.0
514 stars 42 forks source link

Werid behavior on town 3 and 4 #71

Closed JkhwungII closed 9 months ago

JkhwungII commented 9 months ago

Hi, I tired the model i trained on Town 4 and the model deteted an influx of phantom obstacles like in the video. Is it because train too little epoch or too little data. On town 3 model would be stoped by human behind fence

https://github.com/opendilab/InterFuser/assets/53292322/aa9cf68f-202a-429f-8480-c18f1858752c

https://drive.google.com/file/d/1UbEfer95xR4VS1SzkAWQDLZvVglT6IbT/view?usp=drive_link

deepcs233 commented 9 months ago

Hi! It is becasue when when collect the data, we will record all objects near the ego-agent, including the cars on the bridge. Then when we train the model, we will let the model to predict the objects near the ego-agent. However, we do not exclude the objects that the ego-agent can not observe directly. To solve this issue, you can try to collect more data from other towns or delete the related data.

JkhwungII commented 9 months ago

Thanks