AI KIT & CHIP

scroll

Proofly provides a cutting-edge hardware kit designed for researchers and developers to train and optimize deepfake detection models across various deep learning algorithms.

Proofly Hardware Kit

Key Features:

Seamless Integration: Comes with an easy-to-use SDK for web and mobile applications.
Versatile Applications: Ideal for IoT, ICS, and automotive solutions.

Dev KIT

March 2025

Neural Network Model on FPGA:

Hardware-optimized components power deep learning:
Convolution Engine delivers ultra-fast spatial feature extraction with parallelized operations.
ReLU Module applies activation functions to enhance non-linear model capabilities.
Pooling Module reduces spatial dimensions while preserving critical data, lowering computational load.
Fully Connected Layer efficiently maps features to class decisions with on-chip memory buffers.

Image data (pixels):

Uses an AXI Video Frame Buffer and AXI4-Stream Video Interface to handle decoded image storage and flow. This ensures smooth pixel data transfer to preprocessing stages, maintaining performance efficiency.

Cloud Access to chip

April 2025

AI Commercial KIT

June 2025

Pre-processing (resizing, normalization):

This is the quality preparation stage before neural network processing.
Resize Engine scales images to required dimensions while preserving key features.
Normalize Module standardizes pixel values for numerical stability.
Color Space Converter transforms formats (e.g., RGB to YUV) for optimal neural network compatibility.

Post-processing:

Softmax Module generates probabilistic class distributions for interpretable results.
Custom Thresholding Logic applies decision rules, refining output for specific use cases.

Chip first Gen

September 2025

Our
Partners:

Chip first Gen

September 2025

Chip first Gen

September 2025