00:00:00 - Introduction
00:00:41 - CLIP. How to export to ONNX
00:01:05 - CLIP. List of existed ONNX models.
00:01:47 - CLIP. RockChip RKNN export and inference.
00:02:27 - DINOv2. ONNX export. What you should fix.
00:04:28 - DINOv2. RockChip RKNN export and inference.
00:05:49 - Rock PI 3A inference speed
00:06:27 - ONNXRUNTIME. The problem with ARM devices (compiling)
00:07:49 - ONNX inference example
00:08:03 - RPi 3 inference speed for DINOv2
00:08:38 - Other devices. CPU vs. GPU vs. NPU.
Links:
1) CLIP ONNX - github.com/Lednik7/CLIP-ONNX
2) RKNN inference for CLIP - clehaxze.tw/gemlog/2023/07-15...
3) The list of CPU and GPU inference frameworks - qengineering.eu/deep-learning...
4) Issue about DINOv2 export - github.com/facebookresearch/d...
My LinkedIn - / maltsevanton
My Telegram channel - t.me/CVML_team
My e-mail anton@rembrain.ai
Негізгі бет Ғылым және технология One-shot-learing Computer Vision on Embedded devices (RockChip, RPi, etc.)
Пікірлер: 13