MohitAgarwal17/Dog_Emotion_Recognition ? reverse-engineered prompt

Build me a small Python project that can take a photo and tell me if there is a dog in it, then if there is, classify the dog’s emotion as angry, happy, relaxed, or sad. I also want it to explain the result with a simple heatmap so I can see what part of the image influenced the prediction.

Make it work as a clean inference pipeline, with reusable code in a src folder and training or experimentation notebooks if needed. Include a simple way to download the model weights after setup instead of storing them in the repo, and make sure the same image preprocessing is used everywhere so training and prediction stay consistent. If you need to, look up current docs online for the libraries that make the most sense.

I should be able to call one function with an image path and get back whether it is a dog, the emotion, a confidence score, and a short human readable explanation.