Open rose-jinyang opened 4 years ago
It is quite a problem because we assign different ids to similar cars. But I think the training will still work. You can just try it. When tracking, we also use some location information, so it's OK that reid features are not well trained.
Hi I want to know the result of FairMOT in KITTI or DETRAC benchmark for multi-vehicle tracking. The results are quite expected. Thanks
Hello How are you? Excuse me, I think that this project has some critical problems in implementing a vehicle tracking system. I want to know the result of FairMOT in KITTI or DETRAC benchmark for multi-vehicle tracking. Thanks
Sorry, I didn't have experiments on those datasets.
hi @rose-jinyang @ifzhang can you guide me some vehicle dataset for tracking retraining?
@rose-jinyang @HoangTienDuc I have complete a fork to track vehicles(or cars) using UA-DETRAC as training dataset, and the result is fine, please refer to https://github.com/CaptainEven/FairMOTVehicle
Thanks
@CaptainEven Thanks for this, Can you please upload the trained weights?
Hello How are you? I am going to train a model for tracking cars. I have a question about Re-Id. In general, I think that there may be many "same cars" in a same position or different positions. Here "same cars" means that they have same shape, color and brand although they are different objects. In fact, I looked such cases in several vehicle datasets such as KITTI or DETRAC. We assign different ids to these "same cars" to train Re-Id heatmap. In this case, can the model for Re-Id converge? I think that the model may converge in case to train persons rather than cars. But what about training cars? I doubt this point. Thanks