VISION-V1.2-A
A Single-Tier, High-Precision Object Detection Framework
Architecture Overview
VISION12n is a high-performance, single-tier object detection model developed by Vigor Infotech. It utilizes a custom convolutional backbone and a modular detection head optimized for real-time multi-class object detection across diverse environments.
- Backbone built with Conv, C3, and A2C2f modules
- 5-stage spatial reduction pipeline (P1 to P5)
- Fusion of high, mid, and low-level features
- Custom Segment head for final detection
Feature Pyramid Flow
- P1: Conv layer (64 filters)
- P2: Conv layer (128 filters)
- P3: C3 ×2 (256 filters)
- P4: A2C2f ×4 (512 filters)
- P5: A2C2f ×4 (1024 filters)
- Feature pyramid fusion via upsampling and Concat
Detection Head
- Upsampling with nearest neighbor strategy
- Skip connections from backbone layers P3, P4, P5
- Multi-scale detection using Segment module
- Outputs bounding boxes + class predictions
Model Specifications
Total Parameters: ~30M
FLOPs: ~145.1 GFLOPs
Depth: 510 layers
Object Classes: 10
Input Size: 512×512 (recommended)
Modules: Conv, C3k2, A2C2f, Segment
Activation: SiLU
Upsampling: Nearest neighbor (2x)
Framework: PyTorch
System Requirements
CPU
- Minimum: Intel Core i3-2100 / Jetson Nano
- Recommended: Intel Core i5 8th Gen or higher / AMD Ryzen 5 or better
GPU
- Minimum: Not required (runs on CPU)
- Recommended: NVIDIA GTX 1660 / RTX 2060 or better for faster training
RAM
- Minimum: 8 GB
- Recommended: 16 GB or more
Storage
- Minimum: 5 GB free space (HDD supported)
- Recommended: 256 GB SSD or higher
Operating System
- Minimum: Windows 10 / Ubuntu 20.04
- Recommended: Windows 11 / Ubuntu 22.04 LTS
Python Version
- Recommended: Python 3.10
- Supported: 3.8 to 3.11
Frameworks & Libraries
- PyTorch 2.1+, OpenCV, torchvision
- NumPy, SciPy, Matplotlib, Pandas
Optional Hardware (Robotics/Drones)
- Pixhawk, LiDAR, GPS, Ultrasonic Sensors
- Gimbal-stabilized camera, telemetry module
Edge Deployment
- Minimum: Jetson Nano
- Recommended: Jetson Xavier NX
- Raspberry Pi 4 (light inference only)
Cloud & Server Compatibility
- AWS EC2 (g4dn, x2gd), GCP AI Platform
- Vigor IntelliOps Hub
About Vigor Infotech Limited
Vigor Infotech Limited is an Indian IT conglomerate committed to transforming the future through innovation in technology and sustainability. Founded by visionary industrialist Abhishek Tiwari and co-led by Sanjeevani Mishra, the company has grown into a powerhouse of the IT industry.
Corporate Headquarters
Vigor Corporate HeadquartersSector 21, Central Commercial District
Atal Nagar, Near Mantralaya Bhavan
Nava Raipur, Chhattisgarh, India – 492101
