<aside>
⬅️
Back
</aside>
<aside>
Link to the Repo
https://github.com/senthilnathan01/my_ml_projects
</aside>
.jpg)
- Built an art-generating model with neural style transfer (on pre-trained VGG-19) by optimizing costs via Gram matrices.
- Implemented YOLOv2 for car detection in autonomous driving, incorporating IoU and NMS for bounding boxes.
- Engineered FaceNet-inspired recognition system using pre-trained Inception model, triplet loss and L2 distance thresholding.
- Designed U-Net for semantic segmentation with encoder-decoder skips, achieving 80% pixel accuracy for object outlining.
- Constructed ResNet50 with skip connections; trained on hand-sign dataset to 95% accuracy, mitigating vanishing gradients.
- Built binary/multiclass CNN classifiers for emotion detection (99% accuracy) and sign language recognition (95% accuracy).