Some deep learning projects using PyTorch.
- Image segmentation : detecting and delineating each distinct object of interest appearing in an image. The dataset used contains images of humans in different environments and their corresponding masks. The masks can be thought of as the outline of the humans/objects in the image and helps to identify their positions in the image separating them from the background.
- Object Localization : task of locating an instance of a particular object category in an image, typically by specifying a tightly cropped bounding box centered on the instance. The dataset used contains images of objects from 3 categories : mushroom, cucumber and eggplant.
- GradCAM : Gradient-weighted Class Activation Mapping (Grad-CAM), uses the class-specific gradient information flowing into the final convolutional layer of a CNN to produce a coarse localization map of the important regions in the image. In this project, we train a classifier on the dataset of objects from 3 categories : mushroom, cucumber and eggplant. To better understand the model's predictions, we implement a GradCAM function which returns the heatmap of localization map of a given class.
- Aerial Image segmentation : in remote sensing analysis, automatic extraction of road network from satellite or aerial images can be a most needed approach for efficient road database creation, refinement, and updating. Given an aerial image, it is required to output a binary mask for the input image showing for each pixel if it belongs to a road or not. We use a subset of the Massachusetts Roads Dataset which contains aerial images of the state of Massachusetts and their corresponding masks.
- Generative adversarial network : Generative Adversarial Networks (GANs) are a clever way of training a generative model by framing the problem as a supervised learning problem with two sub-models: the generator model that we train to generate new examples, and the discriminator model that tries to classify examples as either real (from the domain) or fake (generated). The two models are trained together in a zero-sum game, adversarial, until the discriminator model is fooled about half the time, meaning the generator model is generating plausible examples.1
Generation of hand-written digits like the MNIST dataset.