chenghuang66 / s2tnet Goto Github PK

Is it the same situation in your training?

As you can see, 16 seconds passed between iter 0/126 and iter 50/126 .

Thank you for your excellent work. Did you conduct any tests to measure the inference latency of your model?

What does each column of the dataset represent in the code?

Evaluation is no longer possible on the Apollo website. We can not get the result by using this algorithm. However, they shared an evaluation file, so, if you have your ground truth then that's possible.

In data_processor.py Line 103~ Line 105: 请问为什么当（abs(relative_cord[0]) > neighbor_distance) | (abs(relative_cord[1]) > neighbor_distance)时，neighbor_mask为True呢？？根据物理意义不应该是当两车距离<= neighbor_distance时，即 (abs(relative_cord[0]) <= neighbor_distance) & (abs(relative_cord[1]) <= neighbor_distance) 时，neighbor_mask才为True吗？