Closed zouzouwei closed 1 year ago
This is the environment where I run.
I also used nuScenes-mini datasets, and fellow you step. but when i run eval.py, i have an error. like this AssertionError: NuScenesMultiView_Map_Dataset2: Database version not found: ./data/nuscenes/v1.0-trainval it seems that dataset = build_from_cfg(cfg, DATASETS, default_args), not deal with mini. how do you do work? thank you!
I find you use cp nuscense_trainval_dataset map/ nuscense_v1.0-mini /map/ whar means,copy nuscense_trainval_dataset to map/ nuscense_v1.0-mini /map/? why do this?
I find you use cp nuscense_trainval_dataset map/ nuscense_v1.0-mini /map/ whar means,copy nuscense_trainval_dataset to map/ nuscense_v1.0-mini /map/? why do this?
because , model will lodaing map/expansion information。
I also used nuScenes-mini datasets, and fellow you step. but when i run eval.py, i have an error. like this AssertionError: NuScenesMultiView_Map_Dataset2: Database version not found: ./data/nuscenes/v1.0-trainval it seems that dataset = build_from_cfg(cfg, DATASETS, default_args), not deal with mini. how do you do work? thank you!
sorry, i forget write this error: mmdet3d/datasets/nuscenes_monocular_dataset_map_2.py L7
self.nusc = NuScenes(version='v1.0-mini', dataroot=self.data_root, verbose=True)
I find you use cp nuscense_trainval_dataset map/ nuscense_v1.0-mini /map/ whar means,copy nuscense_trainval_dataset to map/ nuscense_v1.0-mini /map/? why do this?
because , model will lodaing map/expansion information。
I have an error, like this: No such file or directory: './data/nuscenes/maps/expansion/singapore-hollandvillage.json' Where can I find singapore-hollandvillage.json or how to make it? can I see you catalogue, thanks!
I find you use cp nuscense_trainval_dataset map/ nuscense_v1.0-mini /map/ whar means,copy nuscense_trainval_dataset to map/ nuscense_v1.0-mini /map/? why do this?
because , model will lodaing map/expansion information。
I have an error, like this: No such file or directory: './data/nuscenes/maps/expansion/singapore-hollandvillage.json' Where can I find singapore-hollandvillage.json or how to make it? can I see you catalogue, thanks!
you open this url : https://www.nuscenes.org/nuscenes#download sing up account
find here
Thank you for your reply. I ran tets.py and got the test results. You can try this, I think eval.py no input model.
python tools/test.py configs/fastbev/exp/paper/fastbev_m1_r18_s320x880_v200x200x4_c192_d2_f4.py download_model/epoch_20.pth --eval bbox
Thank you for your reply. I ran tets.py and got the test results. You can try this, I think eval.py no input model. python tools/test.py configs/fastbev/exp/paper/fastbev_m1_r18_s320x880_v200x200x4_c192_d2_f4.py download_model/epoch_20.pth --eval bbox
Thank you for your prompt . I solve this quesation
当您运行Fast-Bev/tools/test.py时,使用M1模型测试在一帧的六张图片需要多长时间?我在T1200(相当于RTX 1080的算力)上进行了测试,测试一帧需要2秒左右。它与论文中的表1和表11的耗时有很大不同。我使用M4模型测试时,每帧也用了4秒左右。 你测试了多久?如果你用其他模型测试它,你也可以详细说明下。 当我训练时,我在两张TITAN RTX TU102(相当于RTX 3080的算力)上训练m1模型,每张卡的批大小为8,使用的nuScenes数据集,训练了21个小时,第一个周期只训练到4530/8007,这训练速度也太慢了吧? 我在T1200(相当于RTX 1080的算力)上训练m1模型,批大小为1,使用的nuScenes-mini数据集,训练了5个小时,只训练到第三个周期的730/1640。训练集只有大约320张照片,为什么这里是1640?
1, dataset nuscenes version v1.0-mini
Modify the code: tools/create_data.py Line 188-212
tools/data_converter/nuscenes_converter.py Line 45
tools/data_converter/nuscenes_seq_converter.py Line 14-16
later, i run python tools/create_data.py nuscenes python tools/data_converter/nuscenes_seq_converter.py ps: cp nuscense_trainval_dataset map/ nuscense_v1.0-mini /map/ Got these:![image](https://user-images.githubusercontent.com/15826704/219848879-505be8f2-922f-4ee6-8e05-0b4e03198325.png)
2.run test.py
python tools/eval.py configs/fastbev/exp/paper/fastbev_m0_r18_s256x704_v200x200x4_c192_d2_f4.py --out temp_out/temp.pkl --eval bbox later:
Why did it turn out like this?
ps: i try run code: python tools/test.py configs/fastbev/exp/paper/fastbev_m0_r18_s256x704_v200x200x4_c192_d2_f4.py pretrained_models/cascade_mask_rcnn_r18_fpn_coco-mstrain_3x_20e_nuim_bbox_mAP_0.5110_segm_mAP_0.4070.pth --eval bbox
result same