MCG-NJU/MixFormer

test on my own dataset

Closed this issue · 2 comments

您好,非常感谢您的工作!我对您的工作很感兴趣,想要测试我自己的数据集时遇到了问题。

我在单卡2080Ti上执行命令 bash tracking/test_mixformer_cvt.sh。 使用MixFormer-1k测试我的Anti-UAV数据集
python tracking/test.py mixformer_cvt_online baseline_1k --dataset anti_uav_train --threads 1 --num_gpus 1 --params__model mixformer_cvt_online_1k.pth.tar --params__search_area_scale 4.55
使用的模型是的mixformer_online_1k.pth.tar,这里我把模型文件名改成了 mixformer_cvt_online_1k.pth.tar

运行后出现Warning报错 Warning: Pretrained CVT weights are not loaded ,但在我的数据集目录下能生成对应的Bbox坐标和time的txt文件(第一个序列没成功生成),而 RESULTS_PATH 和 RESULTS_PLOT_PATH 没有文件。

请问这样的结果合理吗,得到的结果是否可认为是MixFormer-1k模型的结果?

运行结果如下:

{'model': 'mixformer_cvt_online_1k.pth.tar', 'search_area_scale': 4.55, 'max_score_decay': 1.0, 'vis_attn': 0}
tracking/../lib/test/evaluation/anti_uavdataset.py:29: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
ground_truth_rect = np.array(gt_rect_values)
test config: {'MODEL': {'HEAD_TYPE': 'CORNER', 'HIDDEN_DIM': 384, '.....
}}}
search_area_scale: 4.55
Evaluating 1 trackers on 50 sequences
Tracker: mixformer_cvt_online baseline_1k None , Sequence: /root/autodl-tmp/MixFormer/data/Anti-UAV/train/20190925_101846_1_1
Warning: Pretrained CVT weights are not loaded
head channel: 384
Online size is: 3
Update interval is: 200
max score decay = 1.0
can't multiply sequence by non-int of type 'list'
Tracker: mixformer_cvt_online baseline_1k None , Sequence: /root/autodl-tmp/MixFormer/data/Anti-UAV/train/20190925_101846_1_2
Warning: Pretrained CVT weights are not loaded
head channel: 384
Online size is: 3
Update interval is: 200
max score decay = 1.0
Using /root/.cache/torch_extensions as PyTorch extensions root...
Detected CUDA files, patching ldflags
Emitting ninja build file /root/.cache/torch_extensions/_prroi_pooling/build.ninja...
Building extension module _prroi_pooling...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
ninja: no work to do.
Loading extension module _prroi_pooling...
FPS: 23.317997237067637
Tracker: mixformer_cvt_online baseline_1k None , Sequence: /root/autodl-tmp/MixFormer/data/Anti-UAV/train/20190925_101846_1_3
Warning: Pretrained CVT weights are not loaded
head channel: 384
Online size is: 3
Update interval is: 200
....

期待您的答复!

hi,感谢对我们工作的关注,这个warning可以忽略,其实是可以修改一下代码就不显示了,我们后续更新该repo的话会改一下这个问题,详见:


你只需要把合适的测试跟踪模型路径load对就可以了

了解了,我的模型是按照建议放在models文件夹下的,路径是正确的,只是改了模型的名称,因为我百度云下载是mixformer_online_1k.pth.tar,我理解为这个就是mixformer_cvt_online_1k.pth.tar模型。
非常感谢您的答复!