app.py

pip install --upgrade pip
!pip install torch
import streamlit as st
import cv2
import numpy as np
import torch
from PIL import Image
from detectron2.config import get_cfg
from detectron2.projects.deeplab import add_deeplab_config
from detectron2.engine import DefaultPredictor
from mask2former import add_maskformer2_config
from detectron2.utils.colormap import random_color
import os

@st.cache_resource
def setup_config(weights_path):
    cfg = get_cfg()
    cfg.set_new_allowed(True)
    add_deeplab_config(cfg)
    add_maskformer2_config(cfg)
    cfg.merge_from_file("configs/maskformer2_R50_bs16_50ep.yaml")
    cfg.MODEL.WEIGHTS = weights_path
    cfg.MODEL.DEVICE = "cpu"  # Use CPU for inference
    cfg.freeze()
    return cfg

def area(mask):
    if mask.size == 0: return 0
    return np.count_nonzero(mask) / mask.size

def vis_mask(input, mask, mask_color):
    fg = mask > 0.5
    rgb = np.copy(input)
    rgb[fg] = (rgb[fg] * 0.5 + np.array(mask_color) * 0.5).astype(np.uint8)
    return Image.fromarray(rgb)

def show_image(I, pool):
    already_painted = np.zeros(np.array(I).shape[:2])
    input = I.copy()
    for mask in pool:
        already_painted += mask.astype(np.uint8)
        overlap = (already_painted == 2)
        if np.sum(overlap) != 0:
            input = Image.fromarray(overlap[:, :, np.newaxis] * np.copy(I) + np.logical_not(overlap)[:, :, np.newaxis] * np.copy(input))
            already_painted -= overlap
        input = vis_mask(input, mask, random_color(rgb=True))
    return input
import gdown

gdown.download("https://drive.google.com/uc?id=1sCZM5j2pQr34-scSEkgG7VmUaHJc8n4d", "unsam_plus_sa1b_1perc_ckpt_50k.pth", quiet=False)
gdown.download("https://drive.google.com/uc?id=1qUdZ2ELU_5SNTsmx3Q0wSA87u4SebiO4", "unsam_sa1b_4perc_ckpt_200k.pth", quiet=False)

@st.cache_data
def process_image(image, model_type):
    if model_type == "UNSAM+":
        weights_path = "unsam_plus_sa1b_1perc_ckpt_50k.pth"
    else:  # UNSAM
        weights_path = "unsam_sa1b_4perc_ckpt_200k.pth"
    
    cfg = setup_config(weights_path)
    predictor = DefaultPredictor(cfg)
    
    inputs = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2BGR)
    outputs = predictor(inputs)['instances']
    
    masks = []
    for score, mask in zip(outputs.scores, outputs.pred_masks):
        if score < 0.5: continue
        masks.append(mask.cpu().numpy())
    
    sorted_masks = sorted(masks, key=lambda m: area(m), reverse=True)
    result_image = show_image(np.array(image), sorted_masks)
    
    return result_image

st.title("UNSAM and UNSAM+ Image Segmentation")

uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"])

if uploaded_file is not None:
    image = Image.open(uploaded_file)
    
    col1, col2, col3 = st.columns(3)
    
    with col1:
        st.header("Original Image")
        st.image(image, use_column_width=True)
    
    with col2:
        st.header("UNSAM+ Output")
        unsam_plus_output = process_image(image, "UNSAM+")
        st.image(unsam_plus_output, use_column_width=True)
    
    with col3:
        st.header("UNSAM Output")
        unsam_output = process_image(image, "UNSAM")
        st.image(unsam_output, use_column_width=True)

else:
    st.write("Please upload an image to see the segmentation results.")