Justin

Results 21 comments of Justin

Many thanks for replying. Looking forward to this feature can be added. @voldemortX

So if the NMS op is included in the model, the conversion can not work and will enconter error? @voldemortX

Thanks for replying, can you sharing the code to do the visualization as below ?

Many thanks for the explaining, I got it now, :) @lucastabelini @GongDianCheng

Hi, where can I change the pred_bbox to pred_bbox[:, :7]? @Cedarch

Thank you. I have finished the evaluattion, but the results are very bad (model: PV-RCNN++ ResNet): OBJECT_TYPE_TYPE_VEHICLE_LEVEL_1/AP: 0.0436 OBJECT_TYPE_TYPE_VEHICLE_LEVEL_1/APH: 0.0422 OBJECT_TYPE_TYPE_VEHICLE_LEVEL_2/AP: 0.0373 OBJECT_TYPE_TYPE_VEHICLE_LEVEL_2/APH: 0.0361 OBJECT_TYPE_TYPE_PEDESTRIAN_LEVEL_1/AP: 0.0626 OBJECT_TYPE_TYPE_PEDESTRIAN_LEVEL_1/APH: 0.0540 OBJECT_TYPE_TYPE_PEDESTRIAN_LEVEL_2/AP: 0.0520...

Thank you! The number of samples of the processed dataset is 20% of the raw data, so the storage is reduced?