This work solves two typical problems of Deep Learning: Multi-label classification problem and Re-identification problem. The resolution sees the Transfer Learning of one of the most famous CNN ResNet50 on the ImageNet dataset as the primary method. Two different solutions are implemented, one for a single problem and after testing the various hyper-parameters, compared to each other to validate the performance.
The code are in Colab format. To run it on Colab you need to:
-
On Google Drive
- Create a folder called DL in your main drive folder
- Copy the dataset.zip file inside of it
-
On Colab
- Open Multiclassification.ipynb and Re-identification.ipynb files
-
Run them.
The code have a "Setting" cell that provide to connect Colab with your Google Drive, copy the dataset inside the machine and unzip it in order to speed up the runtime.