Can We Use files, codes from this course for any Competition

I want to use the datasets, files and all the weights that are given in the course assignments of transfer learning and yolo algorithms to build my model in Kaggle competition , So is it permissible to use these files? Is it ok If I copy paste the weights , Utils files in Kaggle ?

or is it wrong to use those files for competition ? If It’s wrong then please give me a solution on how to solve the YOLO object detection problem in real time ? any links any guidelines any references will be useful …

thanks in advance …

You’d have to check with Kaggle to see whether you’re allowed to submit someone else’s work as your own in their contests.

Coursera’s course materials are only intended for your use in course study. I don’t think you’re allowed to use it for other purposes.

So what am I supposed to do , I want to do transfer learning for YOLOv4 , How can I do it on coco dataset, any references for implementing it ? I want to implement it in tensorflow .

The code in this course was originally based on YOLO v2, and to the best of my knowledge it still is, so if v4 is your objective it’s not clear how it would help. I don’t think you can take weights trained on v2 and apply them in any way using a v4 model, as the networks are profoundly different. v2 has 19 layers; v4 has 53 layers organized into 4 blocks.

Also to the best of my knowledge, all the YOLO versions are open and you can get the original code from the Internet.

Finally, I concur that taking code from this class and using it in a Kaggle competition is likely to violate the terms of use of Coursera and Kaggle alike.

PS: I’m not clear what modification YOLO would need in order to qualify for ‘real time’. Here is a quote from the 2015 research paper…

Our base network runs at 45 frames per second with no batch processing on a Titan X GPU and a fast version runs at more than 150 fps. This means we can process streaming video in real-time with less than 25 milliseconds of latency.

@sarabhian
Try doing a web search for “kaggle yolo tutorial”.

I got my needs , I just have to clone the github repos of pre trained model and pre trained weights and use it in my code…this is what a standard way prescribed by many…

thank you for your help

Can you clarify if you are using artifacts pre-trained on v2 with a v4 model?

@ai_curious, No I am not using v2 artifacts , I am using tensorflow hub for pre-trained models which are fine tunable, (basically faster RCNN and efficient net)

I also tried to use darknet and darkflow repositories ,yad2k and official tensorflow/models github repo,

but what I got is very low accuracy in my competition :pensive:, getting only 50% accuracy isn’t good thing for a huge dataset , atleast I should get 75% accuracy, I am trying to retrain the model on the given dataset as transfer learning is only able to give maximum 65% output for all the compititors (even experts)