github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #6949] Is there a better model that can accurately recognize image information?下载了好几个多模态的模型，图片识别效果都不好 #4397

New Issue
Closed
opened 2026-04-12 15:20:17 -05:00 by GiteaMirror · 5 comments
GiteaMirror commented 2026-04-12 15:20:17 -05:00
Owner
Copy Link

Originally created by @goactiongo on GitHub (Sep 25, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/6949

What is the issue?

"Using fastgpt --onapi to call local Ollama models, I have downloaded several multimodal models, but the image recognition accuracy is not good. Is there a better model that can accurately recognize image information?"

使用fastgpt--onapi调用ollama本地模型，下载了好几个多模态的模型，图片识别效果都不准确。
有没有好一点的可以识别图片信息的模型

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.3.11

Originally created by @goactiongo on GitHub (Sep 25, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/6949 ### What is the issue? "Using fastgpt --onapi to call local Ollama models, I have downloaded several multimodal models, but the image recognition accuracy is not good. Is there a better model that can accurately recognize image information?" 使用fastgpt--onapi调用ollama本地模型，下载了好几个多模态的模型，图片识别效果都不准确。 有没有好一点的可以识别图片信息的模型 ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.3.11
GiteaMirror added the question label 2026-04-12 15:20:17 -05:00
GiteaMirror commented 2026-04-12 15:20:19 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 25, 2024):

Which models have you tried? The current vision-capable models in the ollama library can be found here.

qwen2-vl (https://github.com/ollama/ollama/issues/6564) and pixtral (https://github.com/ollama/ollama/issues/6748) are apparently quite good but not yet supported in ollama.

<!-- gh-comment-id:2373298323 --> @rick-github commented on GitHub (Sep 25, 2024): Which models have you tried? The current vision-capable models in the ollama library can be found [here](https://ollama.com/search?c=vision). qwen2-vl (https://github.com/ollama/ollama/issues/6564) and pixtral (https://github.com/ollama/ollama/issues/6748) are apparently quite good but not yet supported in ollama.
GiteaMirror commented 2026-04-12 15:20:19 -05:00
Author
Owner
Copy Link

@goactiongo commented on GitHub (Sep 25, 2024):

minicpm-v:8b
llava:13b
bakllava
blackened/llama-3-8b-gpt-4o-ru1.0:latest
gemma2:27b
llava-llama3

The image recognition performance of these models is not accurate, for example, recognizing the information on images of vehicle dashboard displays

<!-- gh-comment-id:2373330445 --> @goactiongo commented on GitHub (Sep 25, 2024): minicpm-v:8b llava:13b bakllava blackened/llama-3-8b-gpt-4o-ru1.0:latest gemma2:27b llava-llama3 The image recognition performance of these models is not accurate, for example, recognizing the information on images of vehicle dashboard displays
GiteaMirror commented 2026-04-12 15:20:19 -05:00
Author
Owner
Copy Link

@olokelo commented on GitHub (Sep 26, 2024):

minicpm-v is the most advanced model available in ollama right now. It's based on qwen2 so it should work fine for Chinese text as well. Make sure your input images are under 1.8mpx as this is the limitation of minicpm-v. Based on my testing it isn't perfect but it's certainly more accurate than older llava based models. I was able to do some OCR tasks with it.

There're also other models that will probably be added pretty soon: pixtral, qwen2-vl and llama 3.2 vision, however for now you can't run them with ollama. Keep an eye on issues listed by @rick-github to know when they're available.

<!-- gh-comment-id:2376887553 --> @olokelo commented on GitHub (Sep 26, 2024): minicpm-v is the most advanced model available in ollama right now. It's based on qwen2 so it should work fine for Chinese text as well. Make sure your input images are under 1.8mpx as this is the limitation of minicpm-v. Based on my testing it isn't perfect but it's certainly more accurate than older llava based models. I was able to do some OCR tasks with it. There're also other models that will probably be added pretty soon: pixtral, qwen2-vl and llama 3.2 vision, however for now you can't run them with ollama. Keep an eye on issues listed by @rick-github to know when they're available.
GiteaMirror commented 2026-04-12 15:20:20 -05:00
Author
Owner
Copy Link

@goactiongo commented on GitHub (Sep 26, 2024):

Thanks，appreciate your replay
 

---原始邮件---
发件人: "Aleksander @.>
发送时间: 2024年9月26日(周四) 晚上9:00
收件人: @.>;
抄送: @.@.>;
主题: Re: [ollama/ollama] Is there a better model that can accurately recognize image information?下载了好几个多模态的模型，图片识别效果都不好 (Issue #6949)

minicpm-v is the most advanced model available in ollama right now. It's based on qwen2 so it should work fine for Chinese text as well. Make sure your input images are under 1.8mpx as this is the limitation of minicpm-v. Based on my testing it isn't perfect but it's certainly more accurate than older llava based models. I was able to do some OCR tasks with it.

There're also other models that will probably be added pretty soon: pixtral, qwen2-vl and llama 3.2 vision, however for now you can't run them with ollama. Keep an eye on issues listed by @rick-github to know when they're available.

—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you authored the thread.Message ID: @.***>

<!-- gh-comment-id:2376936311 --> @goactiongo commented on GitHub (Sep 26, 2024): Thanks，appreciate your replay &nbsp; ---原始邮件--- 发件人: "Aleksander ***@***.***&gt; 发送时间: 2024年9月26日(周四) 晚上9:00 收件人: ***@***.***&gt;; 抄送: ***@***.******@***.***&gt;; 主题: Re: [ollama/ollama] Is there a better model that can accurately recognize image information?下载了好几个多模态的模型，图片识别效果都不好 (Issue #6949) minicpm-v is the most advanced model available in ollama right now. It's based on qwen2 so it should work fine for Chinese text as well. Make sure your input images are under 1.8mpx as this is the limitation of minicpm-v. Based on my testing it isn't perfect but it's certainly more accurate than older llava based models. I was able to do some OCR tasks with it. There're also other models that will probably be added pretty soon: pixtral, qwen2-vl and llama 3.2 vision, however for now you can't run them with ollama. Keep an eye on issues listed by @rick-github to know when they're available. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***&gt;
GiteaMirror commented 2026-04-12 15:20:20 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Oct 24, 2024):

We're seeing good results with llama 3.2. Please give the new 0.4.0 RC a try

https://github.com/ollama/ollama/releases

<!-- gh-comment-id:2434188984 --> @dhiltgen commented on GitHub (Oct 24, 2024): We're seeing good results with llama 3.2. Please give the new 0.4.0 RC a try https://github.com/ollama/ollama/releases
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label question
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#4397
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?