app.py

import gradio as gr
import random
import numpy as np
import os
import requests
import torch
import torchvision.transforms as T
from PIL import Image
from transformers import AutoProcessor, AutoModelForVision2Seq
import cv2
import spaces
import ast

colors = [
    (0, 255, 0),
    (0, 0, 255),
    (255, 255, 0),
    (255, 0, 255),
    (0, 255, 255),
    (114, 128, 250),
    (0, 165, 255),
    (0, 128, 0),
    (144, 238, 144),
    (238, 238, 175),
    (255, 191, 0),
    (0, 128, 0),
    (226, 43, 138),
    (255, 0, 255),
    (0, 215, 255),
    (255, 0, 0),
]

color_map = {
    f"{color_id}": f"#{hex(color[2])[2:].zfill(2)}{hex(color[1])[2:].zfill(2)}{hex(color[0])[2:].zfill(2)}" for color_id, color in enumerate(colors)
}


def is_overlapping(rect1, rect2):
    x1, y1, x2, y2 = rect1
    x3, y3, x4, y4 = rect2
    return not (x2 < x3 or x1 > x4 or y2 < y3 or y1 > y4)

@spaces.GPU
def draw_entity_boxes_on_image(image, entities, show=False, save_path=None, entity_index=-1):
    """_summary_
    Args:
        image (_type_): image or image path
        collect_entity_location (_type_): _description_
    """
    if isinstance(image, Image.Image):
        image_h = image.height
        image_w = image.width
        image = np.array(image)[:, :, [2, 1, 0]]
    elif isinstance(image, str):
        if os.path.exists(image):
            pil_img = Image.open(image).convert("RGB")
            image = np.array(pil_img)[:, :, [2, 1, 0]]
            image_h = pil_img.height
            image_w = pil_img.width
        else:
            raise ValueError(f"invaild image path, {image}")
    elif isinstance(image, torch.Tensor):
        # pdb.set_trace()
        image_tensor = image.cpu()
        reverse_norm_mean = torch.tensor([0.48145466, 0.4578275, 0.40821073])[:, None, None]
        reverse_norm_std = torch.tensor([0.26862954, 0.26130258, 0.27577711])[:, None, None]
        image_tensor = image_tensor * reverse_norm_std + reverse_norm_mean
        pil_img = T.ToPILImage()(image_tensor)
        image_h = pil_img.height
        image_w = pil_img.width
        image = np.array(pil_img)[:, :, [2, 1, 0]]
    else:
        raise ValueError(f"invaild image format, {type(image)} for {image}")

    if len(entities) == 0:
        return image

    indices = list(range(len(entities)))
    if entity_index >= 0:
        indices = [entity_index]

    # Not to show too many bboxes
    entities = entities[:len(color_map)]

    new_image = image.copy()
    previous_bboxes = []
    # size of text
    text_size = 1
    # thickness of text
    text_line = 1  # int(max(1 * min(image_h, image_w) / 512, 1))
    box_line = 3
    (c_width, text_height), _ = cv2.getTextSize("F", cv2.FONT_HERSHEY_COMPLEX, text_size, text_line)
    base_height = int(text_height * 0.675)
    text_offset_original = text_height - base_height
    text_spaces = 3

    # num_bboxes = sum(len(x[-1]) for x in entities)
    used_colors = colors  # random.sample(colors, k=num_bboxes)

    color_id = -1
    for entity_idx, (entity_name, (start, end), bboxes) in enumerate(entities):
        color_id += 1
        if entity_idx not in indices:
            continue
        for bbox_id, (x1_norm, y1_norm, x2_norm, y2_norm) in enumerate(bboxes):
            # if start is None and bbox_id > 0:
            #     color_id += 1
            orig_x1, orig_y1, orig_x2, orig_y2 = int(x1_norm * image_w), int(y1_norm * image_h), int(x2_norm * image_w), int(y2_norm * image_h)

            # draw bbox
            # random color
            color = used_colors[color_id]  # tuple(np.random.randint(0, 255, size=3).tolist())
            new_image = cv2.rectangle(new_image, (orig_x1, orig_y1), (orig_x2, orig_y2), color, box_line)

            l_o, r_o = box_line // 2 + box_line % 2, box_line // 2 + box_line % 2 + 1

            x1 = orig_x1 - l_o
            y1 = orig_y1 - l_o

            if y1 < text_height + text_offset_original + 2 * text_spaces:
                y1 = orig_y1 + r_o + text_height + text_offset_original + 2 * text_spaces
                x1 = orig_x1 + r_o

            # add text background
            (text_width, text_height), _ = cv2.getTextSize(f"  {entity_name}", cv2.FONT_HERSHEY_COMPLEX, text_size, text_line)
            text_bg_x1, text_bg_y1, text_bg_x2, text_bg_y2 = x1, y1 - (text_height + text_offset_original + 2 * text_spaces), x1 + text_width, y1

            for prev_bbox in previous_bboxes:
                while is_overlapping((text_bg_x1, text_bg_y1, text_bg_x2, text_bg_y2), prev_bbox):
                    text_bg_y1 += (text_height + text_offset_original + 2 * text_spaces)
                    text_bg_y2 += (text_height + text_offset_original + 2 * text_spaces)
                    y1 += (text_height + text_offset_original + 2 * text_spaces)

                    if text_bg_y2 >= image_h:
                        text_bg_y1 = max(0, image_h - (text_height + text_offset_original + 2 * text_spaces))
                        text_bg_y2 = image_h
                        y1 = image_h
                        break

            alpha = 0.5
            for i in range(text_bg_y1, text_bg_y2):
                for j in range(text_bg_x1, text_bg_x2):
                    if i < image_h and j < image_w:
                        if j < text_bg_x1 + 1.35 * c_width:
                            # original color
                            bg_color = color
                        else:
                            # white
                            bg_color = [255, 255, 255]
                        new_image[i, j] = (alpha * new_image[i, j] + (1 - alpha) * np.array(bg_color)).astype(np.uint8)

            cv2.putText(
                new_image, f"  {entity_name}", (x1, y1 - text_offset_original - 1 * text_spaces), cv2.FONT_HERSHEY_COMPLEX, text_size, (0, 0, 0), text_line, cv2.LINE_AA
            )
            # previous_locations.append((x1, y1))
            previous_bboxes.append((text_bg_x1, text_bg_y1, text_bg_x2, text_bg_y2))

    pil_image = Image.fromarray(new_image[:, :, [2, 1, 0]])
    if save_path:
        pil_image.save(save_path)
    if show:
        pil_image.show()

    return pil_image




ckpt = "microsoft/kosmos-2-patch14-224"

model = AutoModelForVision2Seq.from_pretrained(ckpt).to("cuda")
processor = AutoProcessor.from_pretrained(ckpt)

@spaces.GPU
def generate_predictions(image_input, text_input, question=None):

    # Save the image and load it again to match the original Kosmos-2 demo.
    # (https://github.com/microsoft/unilm/blob/f4695ed0244a275201fff00bee495f76670fbe70/kosmos-2/demo/gradio_app.py#L345-L346)
    user_image_path = "/tmp/user_input_test_image.jpg"
    image_input.save(user_image_path)
    # This might give different results from the original argument `image_input`
    image_input = Image.open(user_image_path)

    if text_input == "Brief":
        text_input = "<grounding>An image of"
    elif text_input == "Detailed":
        text_input = "<grounding>Describe this image in detail:"
    if question:
        text_input = f"<grounding>{question}"
    print(text_input)
    inputs = processor(text=text_input, images=image_input, return_tensors="pt").to("cuda")

    generated_ids = model.generate(
        pixel_values=inputs["pixel_values"],
        input_ids=inputs["input_ids"],
        attention_mask=inputs["attention_mask"],
        image_embeds=None,
        image_embeds_position_mask=inputs["image_embeds_position_mask"],
        use_cache=True,
        max_new_tokens=128,
    )

    generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]

    # By default, the generated  text is cleanup and the entities are extracted.
    processed_text, entities = processor.post_process_generation(generated_text)

    annotated_image = draw_entity_boxes_on_image(image_input, entities, show=False)

    color_id = -1
    entity_info = []
    filtered_entities = []
    for entity in entities:
        entity_name, (start, end), bboxes = entity
        if start == end:
            # skip bounding bbox without a `phrase` associated
            continue
        color_id += 1
        # for bbox_id, _ in enumerate(bboxes):
            # if start is None and bbox_id > 0:
            #     color_id += 1
        entity_info.append(((start, end), color_id))
        filtered_entities.append(entity)

    colored_text = []
    prev_start = 0
    end = 0
    for idx, ((start, end), color_id) in enumerate(entity_info):
        if start > prev_start:
            colored_text.append((processed_text[prev_start:start], None))
        colored_text.append((processed_text[start:end], f"{color_id}"))
        prev_start = end

    if end < len(processed_text):
        colored_text.append((processed_text[end:len(processed_text)], None))

    return annotated_image, colored_text, str(filtered_entities)

term_of_use = """
### Terms of use  
By using this model, users are required to agree to the following terms:  
The model is intended for academic and research purposes. 
The utilization of the model to create unsuitable material is strictly forbidden and not endorsed by this work. 
The accountability for any improper or unacceptable application of the model rests exclusively with the individuals who generated such content. 

### License
This project is licensed under the [Microsoft Open Source Code of Conduct](https://opensource.microsoft.com/codeofconduct).
"""

with gr.Blocks(title="Kosmos-2", theme=gr.themes.Base()).queue() as demo:
    gr.Markdown(("""
        # Kosmos-2: Grounding Multimodal Large Language Models to the World
        [[Paper]](https://arxiv.org/abs/2306.14824) [[Code]](https://github.com/microsoft/unilm/blob/master/kosmos-2)
        ### This model can answer visual questions, does localize objects in a given image, and even caption the image without hallucination!
        ### To get started, simply pick one of the images. Pick "Brief" or "Detailed" input for captioning. For visual question answering, pick "None" and enter your question.
        """))
    with gr.Row():
        with gr.Column():
            image_input = gr.Image(type="pil", label="Test Image")
            text_input = gr.Radio(["Brief", "Detailed", "None"], label="Captioning Detail", value="Brief")
            question = gr.Textbox(label="Visual Question Answering")
            run_button = gr.Button(value="Run", visible=True)

        with gr.Column():
            image_output = gr.Image(type="pil")
            text_output1 = gr.HighlightedText(
                                label="Generated Description",
                                combine_adjacent=False,
                                show_legend=True,
                            )

    with gr.Row():
        with gr.Column():
            gr.Examples(examples=[
                        ["/content/IMG_4509.jpg", "Detailed", None],
                        ["/content/IMG_4509.jpg", "Brief", None],
                        ["/content/IMG_4509.jpg", "None", "What is in this image?"],
                    ], inputs=[image_input, text_input, question])
        
    gr.Markdown(term_of_use)

    # record which text span (label) is selected
    selected = gr.Number(-1, show_label=False, visible=False)

    # record the current `entities`
    entity_output = gr.Textbox(visible=False)

    # get the current selected span label
    def get_text_span_label(evt: gr.SelectData):
        if evt.value[-1] is None:
            return -1
        return int(evt.value[-1])
    # and set this information to `selected`
    text_output1.select(get_text_span_label, None, selected)

    # update output image when we change the span (enity) selection
    def update_output_image(img_input, image_output, entities, idx):
        entities = ast.literal_eval(entities)
        updated_image = draw_entity_boxes_on_image(img_input, entities, entity_index=idx)
        return updated_image
    selected.change(update_output_image, [image_input, image_output, entity_output, selected], [image_output])

    run_button.click(fn=generate_predictions,
                     inputs=[image_input, text_input, question],
                     outputs=[image_output, text_output1, entity_output],
                     show_progress=True, queue=True)

demo.launch(debug=True)