问题1:UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument.
Solution:
According to the error prompt, click the blue font of the above error prompt, and then jump to the corresponding functional.py file
Enter according to the search ctr+F shortcut key:
_VF.meshgrid
Find the corresponding location, and then find the location shown in the picture below:
Add the following code indexing = ' ij '
Use of torch.meshgrid
torch.meshgrid(*tensors, indexing=None)
indexing is a parameter of torch.meshgrid.
The function of torch.meshgrid is to generate "grid data", such as generating all pixel coordinates of an image.
indexing='ij'
返回的:
grid_i shape: torch.Size([7, 4])
grid_j shape: torch.Size([7, 4])Their shapes are all [W,H]
shape, the 0th dimension is W, and the 1st dimension is H.
Input torch.meshgrid parameter W_arange, the relative order of H_arange is consistentindexing=‘xy’
返回的:
grid_x shape: torch.Size([4, 7])
grid_y shape: torch.Size([4, 7])Their shapes are [H, W]
The first dimension of shape is W, and the 0th dimension is H.
Input the torch.meshgrid parameter W_arange, and the relative order of H_arange is reversed.
Question 2: [mov,mp4,m4a,3gp,3g2,mj2 @ 000001fe939c8d80] moov atom not found
If there is a problem with the video file, you can check whether your video file can still be opened. If it can be opened and the format is .mp4, there is no problem. If it cannot be opened through the player, you can re-record the video file.
Problem 3: The video cannot be saved when running Yolov5_DeepSort target tracking
You can see from the picture below that I have run successfully and the detection has been completed.
When setting the parameters of track.py, put the result file in the output folder:
But in the end there is no output file in my output file
Solution:
Find the track.py file, find the corresponding detect function under the track.py file, and enter in the corresponding position:
save_vide=True