Deep learning and computer vision
Download as PPTX, PDF11 likes2,367 views
The document discusses practical applications of deep learning in various fields such as cancer detection and shoplifting prevention. It outlines techniques such as neural networks, model training, and data augmentation, while emphasizing the importance of understanding business needs and ethical concerns. Additionally, it highlights challenges posed by limited sample sizes and biases in machine learning models.
![[LINEAR REGRESSION]](https://image.slidesharecdn.com/deeplearningandcomputervision-190624084257/85/Deep-learning-and-computer-vision-10-320.jpg)
![[FACENET]
T-SNE
Projection of
128D to 2D](https://image.slidesharecdn.com/deeplearningandcomputervision-190624084257/85/Deep-learning-and-computer-vision-25-320.jpg)
![Chihuahua the movie
[DATA AUGMENTATION]](https://image.slidesharecdn.com/deeplearningandcomputervision-190624084257/85/Deep-learning-and-computer-vision-41-320.jpg)
![[DROPOUT]](https://image.slidesharecdn.com/deeplearningandcomputervision-190624084257/85/Deep-learning-and-computer-vision-43-320.jpg)
Deep learning and computer vision
- 1. DEEP LEARNING IN PRACTICE Tess Ferrandez – Microsoft - @TessFerrandez
- 5. SHOPLIFTING
- 6. DEEP LEARNING What’s so magical about it
- 9. int EstimatePrice(...){ price = 100000 + 67000 * area_in_sqm + 200000 * has_pool + 100000 * new_kitchen + 5000 * neighborhood_quality; return price; } Price = b + w1*area_in_sqm + w2*has_pool + ...
- 13. GIVEN ENOUGH SAMPLES, A NEURAL NETWORK WILL FIND THE PATTERN
- 14. GIVEN ENOUGH SAMPLES, A NEURAL NETWORK WILL FIND THE PATTERN
- 15. SEQUENCESPACE CONVOLUTIONAL NN RECURRENT NNDENSE NN
- 18. [ 0.01949719, 0.09399229, -0.01618082, -0.00876935, 0.03146157, 0.06853894, 0.00096175, -0.06854118, -0.04771797, -0.05296798, 0.02119147, 0.00511259, 0.13726683 INTERMEDIATE REPRESENTATIONEMBEDDINGSECRET CODE
- 19. RECOMMEND A BOOK
- 21. ADULT FICTION (-0.6, 0.4) * (0.7, 0.9) *
- 22. FICTION ADULT MATH REFERENCES US-CENTRIC CHICK LIT FUNNY SCI-FI LAWYERS WOULD BRAD PITT PLAY A CHARACTER IN THE MOVIE? EMBEDDING [ 0.01949719, 0.09399229, -0.01618082, -0.00876935, 0.03146157, 0.06853894, 0.00096175, -0.06854118, -0.04771797, -0.05296798, 0.02119147, 0.00511259, 0.1372668
- 23. WORD EMBEDDINGS
- 24. [ 0.01949719, 0.09399229, -0.01618082, -0.00876935, 0.03146157, 0.06853894, 0.00096175, -0.06854118, -0.04771797, -0.05296798, 0.02119147, 0.00511259, 0.1372668
- 28. encoder decoder
- 29. encoder decoder
- 30. IN PRACTICE The secrets behind the magic
- 37. Time for the Epoch Training data Validation data
- 38. MODEL LOSS ACCURACY BASIC 0.2507 91.05%
- 41. Chihuahua the movie [DATA AUGMENTATION]
- 43. [DROPOUT]
- 45. MODEL LOSS ACCURACY BASIC 0.2507 91.05% AUGMENTATION 0.1988 93.68%
- 49. MODEL LOSS ACCURACY BASIC 0.2507 91.05% AUGMENTATION 0.1988 93.68% TRANSFER LEARN 0.01253 99.47%
- 51. APPLIED MACHINE LEARNING When the magic is gone, and we’re left with Software Engineering
- 52. UNDERSTAND THE BUSINESS NEEDS
- 53. WHAT IS THE PROBLEM? HOW WILL THE MODEL BE USED? / REQUIREMENTS HOW IS IT DONE TODAY? IS IT FEASIBLE? ETHICAL CONCERNS UNDERSTAND THE BUSINESS NEEDS
- 54. UNDERSTAND THE BUSINESS NEEDS MINE CLEAN EXPLORE
- 56. UNDERSTAND THE BUSINESS NEEDS MINE CLEAN EXPLORE ENGINEER MODEL DEPLOY
- 57. LOTS OF LABLED SAMPLES and NO CONSEQUENTIAL DECISIONS SHOTS ON GOAL
- 58. LOUD CROWD GOAL VISIBLE SPEED/DIRECTION PLAYER DENSITY PLAYER POSES
- 59. SCENE CHANGES GOAL IN VIEW NEGATIVE SAMPLING
- 60. 5S VIDEOS - AROUND ACTION NEGATIVE SAMPLES FROM ATTACKS
- 61. VGG EMBEDDINGS
- 64. PEOPLE CLUSTERS - SIZES
- 65. https://github.com/fizyr/keras-retinanet model_path = 'c:/Tess/source/vision_samples/models/resnet50_coco_best_v2.1.0.h5' model = models.load_model(model_path, backbone_name='resnet50’) image_path = 'C:/Tess/source/vision_samples/data/images/basket_image.jpg' image = read_image_bgr(image_path) image = preprocess_image(image) image, scale = resize_image(image) # process image boxes, scores, labels = model.predict_on_batch(np.expand_dims(image, axis=0)) from keras_retinanet import models from keras_retinanet.utils.image import read_image_bgr, preprocess_image, resize_image
- 66. GOAL / NO GOAL
- 68. FOCUSED OPTICAL FLOW ON PLAYERS
- 69. DETECTING CANCER VERY FEW POSITIVE SAMPLES EXTREME ACCURACY NEEDS POTENTIAL FOR BIAS
- 70. HARD TO DIFFERENTIATE ONLY PARTIALLY LABLED EXTREMELY LARGE IMAGES
- 72. CONVEX HULL
- 73. SHOPLIFTING VERY FEW POSITIVE SAMPLES VERY FEW SAMPLES PER ACTION TYPE VERY SENSITIVE TO BIAS
- 74. COVERED FACES ALONE MEN 20-40 HOODIES SHOPLIFTING POSES MEN 20-40 HOODIESCOVERED FACES ALONE
- 75. 12:32:00CHRISTMAS FISH EYE DETECT not PREDICT HUD ARTIFACTS
- 76. NEGATIVE SAMPLES FROM SAME VIDEOS PEOPLE SHOPPING
- 77. POSE DETECTION
- 79. CLASSIFICATION AT THE BOX LEVEL
- 80. A LITTLE DOMAIN KNOWLEDGE GOES A LONG WAY
- 81. KISSKeep it Simple …
- 83. DEEP LEARNING IN PRACTICE Tess Ferrandez – Microsoft - @TessFerrandez