There is a problem with reproducing the results！

[MRYUjg]

I modified the code #68 img_norm_cfg = dict( mean=[103.530, 116.280, 123.675], std=[1.0, 1.0, 1.0], to_rgb=False)
and Completed the testCUDA_VISIBLE_DEVICES=0 python tools/test.py projects/configs/VAD/VAD_tiny_e2e.py ckpts/VAD_tiny.pth --launcher none --eval bbox --tmpdir tmp
But the test results are different from the papers

-------------- Motion Prediction --------------

EPA_car: 0.6403061224489796
EPA_pedestrian: 0.44745575221238937
ADE_car: 0.6471808552742004
ADE_pedestrian: 0.7423169016838074
FDE_car: 0.843453049659729
FDE_pedestrian: 0.8746829628944397
MR_car: 0.10041265474552957
MR_pedestrian: 0.0777988614800759
-------------- Planning --------------
gt_car:7.101449275362318
gt_pedestrian:6.550724637681159
cnt_ade_car:5.615942028985507
cnt_ade_pedestrian:4.22463768115942
cnt_fde_car:5.268115942028985
cnt_fde_pedestrian:3.818840579710145
hit_car:4.739130434782608
hit_pedestrian:3.5217391304347827
fp_car:0.2971014492753623
fp_pedestrian:1.1594202898550725
ADE_car:3.7680349349975586
ADE_pedestrian:3.1584718227386475
FDE_car:4.443408489227295
FDE_pedestrian:3.3402748107910156
MR_car:0.5289855072463768
MR_pedestrian:0.2971014492753623
plan_L2_1s:0.40457220504912833
plan_L2_2s:0.6373244035884441
plan_L2_3s:0.9129885829735439
plan_obj_col_1s:0.0
plan_obj_col_2s:0.0
plan_obj_col_3s:0.0
plan_obj_box_col_1s:0.0
plan_obj_box_col_2s:0.0
plan_obj_box_col_3s:0.0
fut_valid_flag:1.0

I found that it could be an issue with trailer and construction_vehicle in object detection

`Per-class results:
Object Class AP ATE ASE AOE AVE AAE
car 0.488 0.394 0.140 0.077 0.234 0.207
truck 0.301 0.653 0.243 0.086 0.294 0.628
bus 0.274 0.777 0.130 0.148 1.284 0.080
trailer 0.000 1.000 1.000 1.000 1.000 1.000
construction_vehicle 0.000 1.000 1.000 1.000 1.000 1.000
pedestrian 0.424 0.681 0.266 0.397 0.431 0.151
motorcycle 0.247 0.678 0.280 1.388 0.160 0.089
bicycle 0.050 0.612 0.343 0.101 1.301 0.031
traffic_cone 0.580 0.579 0.437 nan nan nan
barrier 0.299 1.152 0.170 0.064 nan nan
Formating results & gts by classes
I can't fix it, please help me！

[MRYUjg]

I used a mini dataset and I generated a new pkl,

maybe because of the mini dataset?

[cyfcyfff]

how to get the file v1.0test and v1.0-trainval? when I unzip the mini dataset there is no this two file