WeVisOne-VL
A lightweight vision-language model open-sourced by Tencent WeChat's vision team, uniformly supporting multiple visual tasks including object detection, OCR, counting, segmentation, and alpha matting.
Timeline (last 90 days)
Tencent WeChat's vision team has open-sourced the WeVisOne-VL vision-language model.
WeVisOne-VL uses the Apache 2.0 license.
WeVisOne-VL offers versions at two parameter scales, 2B and 4B.
WeVisOne-VL is paired with a 9.6-million-parameter pixel decoder.
WeVisOne-VL unifies five types of vision tasks—object detection, OCR, counting, segmentation and alpha matting—into a single model.
WeVisOne-VL drives the output of different vision tasks through natural language instructions.
All Facts (6)
WeVisOne-VL uses the Apache 2.0 license.
50%UnverifiedWeVisOne-VL is paired with a 9.6-million-parameter pixel decoder.
50%UnverifiedWeVisOne-VL unifies five types of vision tasks—object detection, OCR, counting, segmentation and alpha matting—into a single model.
50%UnverifiedWeVisOne-VL drives the output of different vision tasks through natural language instructions.
50%UnverifiedTencent WeChat's vision team has open-sourced the WeVisOne-VL vision-language model.
50%UnverifiedWeVisOne-VL offers versions at two parameter scales, 2B and 4B.
50%