What surprised me
- The hardest part is to get the dataset.
- Tensors, broadcasting, SGD, non-linearity, and Sigmoid (and more) took me a minute to truly get.
AI Playground
AI/ML Projects
AI/ML projects that trained computer vision models from scratch or from pre-trained models. Projects are learning output from following the course Deep Learning for Coders from fastai.
Trained models
Is it a dog or a panda?
LiveBinary image classifier trained on a dataset of dogs and pandas using ResNet-34 with transfer learning. Experienced with DataLoaders, data augmentation, and fine-tuning using the Fast.ai library.
ResNet-34Transfer learningBinary classification
3 vs. 7 Digit Classifier
LiveBinary classifier built from scratch using pure PyTorch on the MNIST dataset. Every layer and training loop was implemented manually. Temperature scaling is applied to the sigmoid output to soften confidence scores.
Built from scratchPure PyTorchMNISTTemperature scaling
Live demo
Is it a dog or a panda?
Upload any image and the model will predict whether it is a dog or a panda.
Upload an image to see the prediction.
Built with
fast.aiPyTorchResNet-34FastAPIHugging Face SpacesNext.js