Classification of different product categories with an additional spam class. Implemented with fine tuning of the base model EfficientNetV2M
- download pre-collected data
- in the Data class, image preprocessing is implemented
- in the Classifier class, the neural network itself is implemented, which will be trained we use multiclass categorical entropy, because we have 72 product categories + 1 spam category
Next, we integrate our model into production by loading it from the .h extension file, preprocessing the newly received data and making predictions on them. Resulting Accuracy: 84%