YOLO-NAS-SAM

Object Detection & Image Segmentation with YOLO-NAS + SAM

Overview

GitHub

This project demonstrates how to perform object detection and image segmentation using YOLO-NAS for object detection and SAM for image segmentation. YOLO-NAS developed by DeciAi is a state-of-the-art object detection model optimized for both accuracy and low-latency inference. SAM, on the other hand, is a powerful segmentation model developed by Meta AI.

Technologies Used

• Python
• OpenCV
• PyTorch
• Numpy
• Matplotlib

YOLO-NAS Overview

YOLO-NAS, short for You Only Look Once with Neural Architecture Search, is a cutting-edge object detection model optimized for both accuracy and low-latency inference. Developed by Deci, YOLO-NAS employs state-of-the-art techniques like Quantization Aware Blocks and selective quantization for superior performance. It sets a new standard for state-of-the-art (SOTA) object detection, making it an ideal choice for a wide range of applications including autonomous vehicles, robotics, and video analytics.

Unique Features of YOLO-NAS

• Utilizes Quantization Aware Blocks for efficient inference without sacrificing accuracy.
• Incorporates AutoNAC technology for optimal architecture design, balancing accuracy, speed, and complexity.
• Supports INT8 quantization for unprecedented runtime performance.
• Employs a hybrid quantization method that selectively quantizes certain parts of the model, reducing information loss and balancing latency and accuracy.
• Pre-training regimen includes automatically labeled data, self-distillation, and large datasets.
• Available under an open-source license with pre-trained weights for research use on SuperGradients, Deci’s PyTorch-based computer vision training library.

YOLO-NAS architecture

• Quantization-aware blocks: These blocks are designed to be quantized efficiently, which can improve performance and reduce latency.
• Selective quantization: This technique allows for quantization of specific layers of the model without sacrificing accuracy.
• Attention mechanism: This mechanism allows the model to focus on the most important parts of an image, which can improve accuracy and reduce inference time.

Training Details

YOLO-NAS undergoes a multi-phase training process that includes pre-training on Object365, COCO Pseudo-Labeled data, Knowledge Distillation (KD), and Distribution Focal Loss (DFL). The model is meticulously trained on Objects365, a comprehensive dataset with 2 million images and 365 categories, for 25-40 epochs, ensuring robust performance.

SAM Overview

SAM (Segment Anything Model) is a large language model from Meta AI that can be used to segment objects in images with high accuracy. SAM is trained on a massive dataset of images and segmentation masks, and it can be used to generate masks for all objects in an image, or for specific objects or regions of interest.

Supported Prompts

• Foreground/background points
• Bounding box
• Mask

Note: Text prompts are explored in the research paper but the capability is not released.

Model Structure

• A ViT-H image encoder that runs once per image and outputs an image embedding.
• A prompt encoder that embeds input prompts such as clicks or boxes.
• A lightweight transformer-based mask decoder that predicts object masks from the image embedding and prompt embeddings.

Model Size

• The image encoder has 632M parameters.
• The prompt encoder and mask decoder have 4M parameters.

Training Data

The model was trained on the SA-1B dataset.

Efficient & Flexible Design

SAM is designed to be efficient enough to power its data engine. It is decoupled into a one-time image encoder and a lightweight mask decoder that can run in a web browser in just a few milliseconds per prompt.

Note: SAM is still under development, but it has the potential to revolutionize the way we interact with images and videos.