A presentation at AI For Health in in Paris, France by Eléa PETTON
What if AI was the solution to understand sign language? Eléa Petton
Présentation Machine Learning Engineer AI Solution team @EleaPetton eleapttn Eléa PETTON 2
OVHcloud French & European leader in Cloud Data privacy and sovereignty Not subject to the Cloud Act HDS certification (ISO 27001 norm) OVHcloud certifications Collaboration with public sector 3
OVHcloud values Trust Working Together Passion Disruption Responsibility 4
OVHcloud 5
Public Cloud - AI Tools NOTEBOOK AI Notebooks AI Training AI Deploy 6
Computer Vision 7
Object Detection DOG 0.98 8
YOLOv7 You Only Look Once Family of algorithms Recognition and detection of various objects in pictures or videos Real-time Fast + accurate 9
ASL alphabet detection W 0.89 YOLOv7 custom model WEBCAM real time detection RESULTS saving and displaying 10
Journey to ASL detection ASL ALPHABET database OBJECT STORAGE exportation STREAMLIT APP development YOLOv7 model CUSTOM MODEL test AI APP deployment 100 EPOCHS training 11
AI Notebooks 12
AI Deploy 13
CONCLUSION Train a model with GPU power Store easily and securely the AI model Build and an AI app using the trained AI model Deploy an AI Deploy app with security and high accessibility 14
Thanks! Access to the app for real-time detection 15
Any questions? … 16
References OVHcloud AI examples repository: https://github.com/ovh/ai-training-examples OVHcloud AI documentation: https://docs.ovh.com/gb/en/publiccloud/ai/ YOLOv7 repository: https://github.com/WongKinYiu/yolov7 ASL alphabet dataset: https://public.roboflow.com/object-detection/american-sign-language-letters Full code: https://github.com/eleapttn/yolov7_streamlit_asl_recognition Public cloud access: https://www.ovhcloud.com/fr/public-cloud/ OVHcloud security certifications: https://corporate.ovhcloud.com/fr/trusted-cloud/security-certifications/
In the field of Artificial Intelligence, we often talk about Computer Vision and more particularly Object Detection. But how can this technology help us in daily life and medical field?
During this masterclass, we will see how to train an object detection model (YOLOv7) to recognize the different signs of the ASL (American Sign Language) alphabet through a Jupyter notebook.
Training a model is good, using it is better!
We will use our artificial intelligence model to create an application for real-time detection. Docker technology will then be used to deploy the application in the Cloud and make it accessible.
It will allow anyone to understand sign language with real-time detection and written transcription.
for free. You
can too.