Nb. days
3
Face-to-face?
Yes
Remote?
Yes
Min. participants
2
Max. participants
8
Place
Toulouse, Paris, etc.
Pricing
2400€ HT
Since 2012, Deep Learning has revolutionized the field of computer vision. At the annual ImageNet Large Scale Visual Recognition Challenge (ILSVRC), whose goal is to accurately detect and classify objects and scenes in natural images, the use of Deep Learning helped lower the classification error rate from 25% to 16%. This was only the beginning of this revolution as the following two years saw the error rate drop drastically to a few percent.
You already know the basic principles of Deep Learning and you have already implemented different neural network architectures, but you want to know more about the opportunities offered by Deep Learning in the field of computer vision, this training is for you.
Goals
- Implement efficient preprocessing of an image dataset Increase the image dataset
- Reuse existing models with transfer learning
- Master the different neural network architectures allowing to perform object detection, semantic segmentation, object tracking and action recognition
- Implement concrete cases for each type of network
- Run calculations on CPUs, GPUs and TPUs
- Measure the relevance of the implemented models & Visualize the learning
- Deploy a model in production
To know the dates of the next training sessions,
to adapt this program to your needs
or to obtain additional information,
contact us!
to adapt this program to your needs
or to obtain additional information,
contact us!
Teaching method
- This training is 70% practical and uses illustrated and didactic exercises.
- A daily evaluation of the acquisition of the knowledge of the day before is carried out.
- A synthesis is proposed at the end of the training.
- An evaluation will be proposed to the trainee at the end of the course.
- A course support will be given to each participant including slides on the theory and exercises.
- A sign-in sheet for each half-day of attendance is provided at the end of the course as well as a certificate of completion if the trainee has attended the entire session.
- A follow-up and an exchange with the participants will take place a few days after the training.
Prerequisite
- Trainees must have experience in Deep Learning or have followed the Workshop Deep Learning training (mastery of Machine Learning concepts and have implemented different neural network architectures: convolutional, recurrent, ...)
- Knowledge of Python, scientific libraries (scikit learn, pandas, numpy) and the Deep Learning framework TensorFlow is required.
Targets
- Data Analysts
- Data Scientists
- Developers
Evaluation
This training does not require a formal assessment of learningProgram
Images Classification
- Creation of an image dataset
- Images pre-processing
- Images Augmentations
- Convolution and Pooling
- CNN reference architectures: LeNet, AlexNet, Inception, VGG, ResNet, Xception, SENet, etc.
- Transfer learning: use of pre-trained models
- Classification and Localisation
Hands on: Images Classification
Object Detection
- Sliding window detectors
- Region Proposal Networks (RPN) : R-CNN, Fast R-CNN, Faster R-CNN
- Single Shot detector : YOLO, RetinaNet
Hands on: Object Detection through transfer learning
Semantic & Instance Segmentation
- Fully Convolutional
- Instance Segmentation : Mask R-CNN
- Downsampling et Upsampling
- Semantic Segmentation : U-Net
Hands on: Development of a semantic segmentation model
Transformers
- Sequence to sequence learning
- Attention
- Transformer for natural language processing
- Transformer for computer vision
Hands on: Model Vision Transformer – ViT
Object tracking and action recognition
- Optical Flow : FlowNet & RAFT
- Pose Estimation : PoseNet & MoveNet
Hands on: Development of an action recognition model
Les modèles génératifs
- Deep Generative Models
- Auto Encoders
- Generative Adversarial Network (GAN)
Hands on: Development of a GAN model
Style Transfer
- StyleNet
Hands on: Apply the style of a painting to an image
