Cannot load image when detecting multiple images

[joy-eveline]

I built a custom object detector using tiny Yolov4 on Google Colab. Now I am trying to detect multiple images and save the prediction results to a txt file.

I'm using this command line: ! ./darknet detector test data/obj.data cfg/yolov4-tiny-obj.cfg /mydrive/tiny-yolov4/backup/yolov4-tiny-obj_best.weights -dont_show -ext_output < /mydrive/img.txt > result.txt

This is my input img.txt:

Whe I run this, it tells me it can't load the images:

I have already tried changing from .jpg to .png, made sure there are no empty lines, saved in utf-8 format.

If I run detection on each image seperately, it works.

Does anyone have a solution for this? Thanks.

[ShaneXiangH]

I have the same problem..

Does anyone have a solution for this?

[stephanecharette]

I don't know if this helps, but the DarkHelp wrapper has a --json flag where it will work on multiple images/videos and output a block of JSON text at the end. You can also use --keep to indicate you want it to annotate and keep all the images.

https://www.ccoderun.ca/darkhelp/api/ShellScripting.html

[ShaneXiangH]

Let me describe my problem in detail.

I just follow the steps in the Google Colab the wiki mentions to do the Compilation and Detection, see in this link: https://colab.research.google.com/drive/12QusaaRj_lUwCGDvQNfICpa7kA7_a2dE#scrollTo=ifqddnZ0yjCd

They work well until I enter into the "step 9: Multiple Images at Once" part. This instruction in that part cannot work well:

Note that I changed the “path to image” parameter, since in my Google drive, the path is arranged in this structure:

And in the images.txt, I followed the rules mentioned in the Colab and wrote like this:

However, when I execute the instruction mentioned above, the output shows that it cannot load the images:

And the json file I got by running the instuction:

is just:

I don't know where is the problem, since I didn't change any code or instructions on that Colab except the “path to image” parameter to access the files in my google drive. Does anyone have a solution for this?

[stephanecharette]

I've never used colab. But from that JSON file you posted, my first guess would be that you have \r\n line endings and it is expecting \n only.

[ShaneXiangH]

Ohh you are right, that works for me! Thank you!