search

sean-mcclure / defect_classifier

32 stars 35 forks source link

Where is the extracted_defects.zip file ? #2

Open PradyumnaGupta opened 3 years ago

PradyumnaGupta commented 3 years ago

Hello there , may I know where are the extracted images to which the names in the csv file refer to ? I am testing out some image segmentation models for this task and would really appreciate if you could provide me the images you labeled as I don't really have the time to annotate the images myself . Thanks a lot .

sean-mcclure commented 3 years ago

You have to upload the labelled images from the Turker to the CSV folder, as described in the Medium article.

stikoson commented 3 years ago

Hi Sean, thanks for the great project. I think the question above is that how can we avoid using the Turker and have a dataset with both the extracted_defects and image_labels.csv available right away for training. It happens to me that the number of files in the all_extracts folder is 1403. The image_labels.csv that comes in the repo by default has extract_1508.png. I guess there will be differences in defect extraction depending on which platform and which version of the packages one uses. It probably impacts which images will be skipped.

Hence, I'd like to kindly ask you to publish the extracts that corresponds to the image_labels.csv that comes with you repo. It'd save a great amount of time for us...unless you want everyone to do the turking :) I'm attaching my notebook to see where I got an error due to the differences in my defect extracts vs your image_labels.csv. PCB_Defect_Extraction_and_Classification.zip

zee9999 commented 2 years ago

Hi Sean, thanks for the great project. I think the question above is that how can we avoid using the Turker and have a dataset with both the extracted_defects and image_labels.csv available right away for training. It happens to me that the number of files in the all_extracts folder is 1403. The image_labels.csv that comes in the repo by default has extract_1508.png. I guess there will be differences in defect extraction depending on which platform and which version of the packages one uses. It probably impacts which images will be skipped.

Hence, I'd like to kindly ask you to publish the extracts that corresponds to the image_labels.csv that comes with you repo. It'd save a great amount of time for us...unless you want everyone to do the turking :) I'm attaching my notebook to see where I got an error due to the differences in my defect extracts vs your image_labels.csv. PCB_Defect_Extraction_and_Classification.zip

Hi sean It seems that i have encountred the same error, it would be a great help if u could solve it or give any pointers.

Thanks in advance.

Ankitgun commented 2 years ago

I am having the same issue. I am not sure what's the solution is. I tried to upload the labelled images from the Turker to the CSV folder but it didn't solve the issue.

sean-mcclure commented 2 years ago

Hello. Unfortunately I do not have these images saved. As explained in the article, if you do a few hundred at a time it's not too bad. You can choose how many total images to use in the training.

https://towardsdatascience.com/building-an-end-to-end-deep-learning-defect-classifier-application-for-printed-circuit-board-pcb-6361b3a76232#953a

screenshot