This is an official implementation for "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows".

Official PyTorch implementation of A-ViT: Adaptive Tokens for Efficient Vision Transformer (CVPR 2022)

High-Performance Sparse Linear Algebra on HBM-Equipped FPGAs Using HLS

SSR: Spatial Sequential Hybrid Architecture for Latency Throughput Tradeoff in Transformer Acceleration (Full Paper Accepted in FPGA'24)

FPGA based Vision Transformer accelerator (Harvard CS205)

Simple examples for FPGA design using Vivado HLS for high level synthesis and Vivado for bitstream generation.

PYNQ support and examples for Kria SOMs

[ICCV 2023] I-ViT: Integer-only Quantization for Efficient Vision Transformer Inference

An implementation of SIFT on GPU with OpenCL

Implementation of SIFT Algorithm on FPGA

An FPGA Accelerator for Transformer Inference

Sub part of the SIFT algorithm as a Vitis HLS accelerated kernel

ezSIFT: An easy-to-use standalone SIFT library written in C/C

Machine learning on FPGAs using HLS

Real Time SIFT implementation on FPGA

Using Verilog to implement the SIFT algorithm into an FPGA for small robotic situations

Post-Training Quantization for Vision transformers.

[HPCA 2023] ViTCoD: Vision Transformer Acceleration via Dedicated Algorithm and Accelerator Co-Design

A Comprehensive Dataflow Compiler for High-Level Synthesis

You can run it on pynq z1. The repository contains the relevant Verilog code, Vivado configuration and C code for sdk testing. The size of the systolic array can be changed, now it is 16X16.

Segment Anything combined with CLIP

The repository provides code for running inference with the SegmentAnything Model (SAM), links for downloading the trained model checkpoints, and example notebooks that show how to use the model.

[IJCAI 2022] FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformer

深度学习经典、新论文逐段精读

CLIP (Contrastive Language-Image Pretraining), Predict the most relevant text snippet given an image

XLS: Accelerated HW Synthesis

Kria Vitis platforms and overlays

Edge-MoE: Memory-Efficient Multi-Task Vision Transformer Architecture with Task-level Sparsity via Mixture-of-Experts