Deep Learning for Images with PyTorch
DataCamp
A hands-on course covering four workhorse image tasks in PyTorch. It opens with classification, binary and multi-class, with convolutional neural networks — including how to use transfer learning, loading a pre-trained model and putting ResNet to work on images. It then moves to object recognition, where the model predicts bounding boxes as well as labels, and you compute intersection over union and non-maximum suppression yourself to judge and clean up the results. The third chapter is segmentation — classifying an image pixel by pixel — applying a pre-trained Mask R-CNN and then building a U-Net layer by layer. The last builds a generative adversarial network (GAN), two networks trained against each other, to produce new images. Four chapters, 16 videos, and 58 exercises across four hours, taught by machine-learning engineer Michał Oleszak. It is the third in DataCamp's PyTorch sequence.
What you'll learn
- Binary and multi-class image classification with convolutional neural networks
- Working with pre-trained models: saving and loading, transfer learning, and image classification with ResNet
- Object recognition: predicting bounding boxes alongside labels, up to Faster R-CNN
- Evaluating and filtering detections with intersection over union and non-maximum suppression
- Semantic, instance, and panoptic segmentation, with Mask R-CNN and a U-Net you build yourself
- Generating new images with a generative adversarial network (GAN)
Frequently asked questions about Deep Learning for Images with PyTorch
Who is Deep Learning for Images with PyTorch for?
Developers already comfortable with PyTorch who want image classification, object detection, segmentation, and generation covered in one practical, project-heavy course.
Is Deep Learning for Images with PyTorch free?
No — Deep Learning for Images with PyTorch is a paid course.
What are the prerequisites for Deep Learning for Images with PyTorch?
DataCamp's Introduction to Deep Learning with PyTorch and Intermediate Deep Learning with PyTorch.
Do you need to code for Deep Learning for Images with PyTorch?
Yes — Deep Learning for Images with PyTorch involves hands-on coding.
Does Deep Learning for Images with PyTorch offer a certificate?
Yes. DataCamp Statement of Accomplishment on completion (requires DataCamp Premium).
Why we suggest this course
For a developer who already knows PyTorch and wants the main vision tasks — classification, detection, segmentation, generation — covered in one course. It works the way real projects do: you take pre-trained models off the shelf where the field already has good ones, loading ResNet for classification and Mask R-CNN for instance segmentation, and you build architectures from their parts where building is what teaches you something — a U-Net layer by layer, and both halves of a deep convolutional GAN. The detection metrics you calculate by hand, which is where the intuition comes from. One thing to know: it assumes both of DataCamp's earlier PyTorch courses, so it is a poor entry point if the framework is new. You can sample the opening chapter before subscribing; the full course and its Statement of Accomplishment are part of DataCamp Premium.