Skip to content

Latest commit

 

History

History
48 lines (38 loc) · 1.66 KB

inference.md

File metadata and controls

48 lines (38 loc) · 1.66 KB

Inference Guide

GeoPixel-7B is a powerful model for pixel-grounded analysis of images. Follow the steps below to interact with it via chat.py for detailed descriptions and segmentation mask generation.

Grounded Remote Sensing Image Analysis

To chat with GeoPixel-7B to obtain detailed pixel grounded analysis of image use following command:

CUDA_VISIBLE_DEVICES=0 python chat.py --version='MBZUAI/GeoPixel-7B'

Input the text prompt and then the image path. For example，

Example Usage

- Please input your prompt: Can you provide a thorough description of this image? Please output with interleaved segmentation masks for the corresponding phrases.
- Please input the image path: images/example1.png

Referring Remote Sensing Image Segmentation (RRSIS)

To obtain grounded segmentation masks for specific referring phrases in remote sensing images, use GeoPixel-7B-RES:

CUDA_VISIBLE_DEVICES=0 python chat.py --version='MBZUAI/GeoPixel-7B-RES'

Usage Instructions

Input a text prompt specifying the referring phrase ({ref_exp}) in

Could you provide a segmentation mask for {ref_exp} in this image?

and provide the path to the image.

Example Usage

- Please input your prompt: Could you provide a segmentation mask for {ref_exp} in this image?
- Please input the image path: images/example1-RES.jpg

ref_exp examples for images/example1-RES.jpg:

  • "red car",
  • "white car at right"
  • "tree at bottom right"
  • "a vehicle on the left"
  • "house at bottom right"
  • "purple tree at bottom"
- Please input your prompt: Could you provide a segmentation mask for red car in this image?
- Please input the image path: images/example1-RES.jpg