A practical guide to deep learning
summary
Walks through building real systems: uses Keras with convolutional layers and VGG embeddings per frame to classify five‑second soccer clips, combines audio features and optical flow and achieves >90% shot-on-goal accuracy; shows Chihuahua vs muffin prototype that trains at 98.6% but validates at 91% and improves to ~93% with data augmentation and dropout; solves cancer-slide detection without deep nets via color decomposition and convex‑hull clustering for high precision; for shoplifting uses OpenCV background subtraction, pretrained detectors and pose features to reduce bias and false positives.
AI-written summary · about
chapters (23)
- 0:00Start
- 1:44Cancer Detection
- 2:50Background of Deep Learning
- 5:52Linear Regression
- 8:42Problems of Deep Learning
- 11:06Recurrent Neural Networks
- 13:57A Convolutional Neural Network
- 14:16Pooling
- 15:15Intermediate Layer
- 24:14Generative Network
- 27:31Neural Network
- 29:56Overfitting
- 33:03Learning from Our Past Experiences
- 34:50Transfer Learning
- 43:49Player Density
- 49:44Object Detection
- 50:26Scene Change Detection
- 50:56Optical Flow
- 51:24Focused Optical Flow
- 55:35Shoplifting Case
- 57:54Negative Sampling
- 58:32Background Subtraction
- 1:00:01Keep It Simple
related talks
01h 01mA practical guide to deep learning
35mDeconstructing Deep Learning
32mYou Can Do Deep Learning!
01h 29mComputer Vision: An (Un?)Expected Journey, with Keras and Tensorflow
01h 00mA gentle introduction to Deep Learning
35mFootball video analysis using Deep Learning
01h 04mWhat you always wanted to know about Deep Learning, but were afraid to ask
58mDeep Learning in Real Life