Hi there!
I'll be posting some projects about Machine Learning, Data Science, NLP and AI in general.
- To face this unsupervised learning problem, I've used a dataset that contains 5000 unlabeled images.
- I've used the K-Means algorithm to cluster the N dominant colors of the images.
- Finally, we use a threshold, in the HSV color space, for the color that we want to detect from all the clothes.
- If any of the N dominant colors is inside this threshold, it's classified as this color.
- The dataset contains 10000 samples of reviews and ratings about a fashion e-commerce.
- The reviews have been preprocessed using NLTK.
- BERT has been used for fine-tuning training.
- Summarizes automatically any text or web page of any size using transformers and other models.
- API deployed using Streamlit (Summarizer)
- Transformers used: BART, T5, PEGASUS, Longformer and more.
- Analyzed what purchasing categories have been most affected by Covid (positively and negatively).
- Trained the model using Logistic Regression
- Created some data visualizations
- Got 97,1% of accuracy
- Trained the model using SVM
- Used the Linear, RBF and Polynomial kernel
- Got 79% of accuracy using RBF kernel