OVD_SOSP-B_Internvl_model2

import torch
import requests
from io import BytesIO
from PIL import Image, ImageDraw
from transformers import AutoConfig, AutoModel, AutoTokenizer

repo_id = "xpuenabler/OVD_SOSP-B_Internvl_model2"
image_source = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg"
query = "dog"

# Load image
if image_source.startswith(("http://", "https://")):
    response = requests.get(image_source)
    pil = Image.open(BytesIO(response.content)).convert("RGB")
else:
    pil = Image.open(image_source).convert("RGB")

# Load model
cfg = AutoConfig.from_pretrained(repo_id, trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(cfg.vlm_model_name, trust_remote_code=True, use_fast=False)
model = AutoModel.from_pretrained(repo_id, trust_remote_code=True)
model.eval()

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = model.to(device)

# Run inference
outputs = model.infer_image(image=pil, query=query, tokenizer=tokenizer)
pred_boxes = outputs.pred_boxes[0].float().cpu()

# Visualize
w, h = pil.size
draw = ImageDraw.Draw(pil)
x1, y1, x2, y2 = pred_boxes[0].tolist()
x1, y1, x2, y2 = x1 * w, y1 * h, x2 * w, y2 * h
draw.rectangle([x1, y1, x2, y2], outline="red", width=3)
draw.text((x1, max(y1 - 20, 0)), query, fill="red")

pil.save("output.jpg")
print(f"Saved: output.jpg")

import torch
import requests
from io import BytesIO
from PIL import Image, ImageDraw
from transformers import AutoConfig, AutoModel, AutoTokenizer

repo_id = "xpuenabler/OVD_SOSP-B_Internvl_model2"
image_source = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg"
queries = ["person", "dog"]

# Load image
if image_source.startswith(("http://", "https://")):
    response = requests.get(image_source)
    pil = Image.open(BytesIO(response.content)).convert("RGB")
else:
    pil = Image.open(image_source).convert("RGB")

# Load model
cfg = AutoConfig.from_pretrained(repo_id, trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(cfg.vlm_model_name, trust_remote_code=True, use_fast=False)
model = AutoModel.from_pretrained(repo_id, trust_remote_code=True)
model.eval()

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = model.to(device)

# Run inference
outputs = model.infer_batch(image=pil, queries=queries, tokenizer=tokenizer)
pred_boxes = outputs.pred_boxes.float().cpu()

# Visualize
w, h = pil.size
draw = ImageDraw.Draw(pil)
for boxes, query in zip(pred_boxes, queries):
    x1, y1, x2, y2 = boxes[0].tolist()
    x1, y1, x2, y2 = x1 * w, y1 * h, x2 * w, y2 * h
    draw.rectangle([x1, y1, x2, y2], outline="red", width=3)
    draw.text((x1, max(y1 - 20, 0)), query, fill="red")

pil.save("output.jpg")
print(f"Saved: output.jpg")