app.py

from transformers import  PaliGemmaProcessor, PaliGemmaForConditionalGeneration
from typing import List
import os
import supervision as sv
import uuid
from tqdm import tqdm
import gradio as gr
import torch
from PIL import Image
import spaces
import flax.linen as nn
import jax
import string
import functools
import jax.numpy as jnp
import numpy as np
import re


device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model_id = "google/paligemma-3b-mix-448"
model = PaliGemmaForConditionalGeneration.from_pretrained(model_id).eval().to(device)
processor = PaliGemmaProcessor.from_pretrained(model_id)

BOUNDING_BOX_ANNOTATOR = sv.BoundingBoxAnnotator()
MASK_ANNOTATOR = sv.MaskAnnotator()
LABEL_ANNOTATOR = sv.LabelAnnotator()


def calculate_end_frame_index(source_video_path):
    video_info = sv.VideoInfo.from_video_path(source_video_path)
    return min(
        video_info.total_frames,
        video_info.fps * 2
    )


def annotate_image(
    input_image,
    detections,
    labels
) -> np.ndarray:
    output_image = MASK_ANNOTATOR.annotate(input_image, detections)
    output_image = BOUNDING_BOX_ANNOTATOR.annotate(output_image, detections)
    output_image = LABEL_ANNOTATOR.annotate(output_image, detections, labels=labels)
    return output_image

@spaces.GPU
def process_video(
    input_video,
    labels,
    progress=gr.Progress(track_tqdm=True)
):
    video_info = sv.VideoInfo.from_video_path(input_video)
    total = calculate_end_frame_index(input_video)
    frame_generator = sv.get_video_frames_generator(
        source_path=input_video,
        end=total
    )

    result_file_name = f"{uuid.uuid4()}.mp4"
    result_file_path = os.path.join("./", result_file_name)
    with sv.VideoSink(result_file_path, video_info=video_info) as sink:
        for _ in tqdm(range(total), desc="Processing video.."):
            frame = next(frame_generator)
            # list of dict of {"box": box, "mask":mask, "score":score, "label":label}
            results, input_list = parse_detection(frame, labels)
            detections = sv.Detections.from_transformers(results[0])
            final_labels = []
            
            for id in results[0]["labels"]:
              final_labels.append(input_list[id])
            frame = annotate_image(
                input_image=frame,
                detections=detections,
                labels=final_labels,
            )
            sink.write_frame(frame)
    return result_file_path

@spaces.GPU
def infer(
    image: Image.Image,
    text: str,
    max_new_tokens: int
) -> str:
    inputs = processor(text=text, images=image, return_tensors="pt").to(device)
    with torch.inference_mode():
      generated_ids = model.generate(
          **inputs,
          max_new_tokens=max_new_tokens,
          do_sample=False
      )
    result = processor.batch_decode(generated_ids, skip_special_tokens=True)
    return result[0][len(text):].lstrip("\n")

def parse_detection(input_image, input_text):
  prompt = f"detect {input_text}"
  out = infer(input_image, prompt, max_new_tokens=100)
  objs = extract_objs(out.lstrip("\n"), input_image.shape[0], input_image.shape[1], unique_labels=True)
  
  labels = list(obj.get('name') for obj in objs if obj.get('name'))
  print("labels", labels)
  input_list = input_text.split(";")
  for ind, input in enumerate(input_list):
      input_list[ind] = remove_special_characters(input).lstrip("\n").rstrip("\n")
  label_indices = []
  for label in labels:
      label = remove_special_characters(label)
      label_indices.append(input_list.index(label))
  label_indices = torch.tensor(label_indices).to("cuda")
  boxes = torch.tensor([list(obj["xyxy"]) for obj in objs])
  return [{"boxes": boxes, "scores":torch.tensor([0.99 for _ in range(len(boxes))]).to("cuda"), "labels":label_indices}], input_list

_MODEL_PATH = 'vae-oid.npz'

_SEGMENT_DETECT_RE = re.compile(
    r'(.*?)' +
    r'<loc(\d{4})>' * 4 + r'\s*' +
    '(?:%s)?' % (r'<seg(\d{3})>' * 16) +
    r'\s*([^;<>]+)? ?(?:; )?',
)





def _quantized_values_from_codebook_indices(codebook_indices, embeddings):
  batch_size, num_tokens = codebook_indices.shape
  assert num_tokens == 16, codebook_indices.shape
  unused_num_embeddings, embedding_dim = embeddings.shape

  encodings = jnp.take(embeddings, codebook_indices.reshape((-1)), axis=0)
  encodings = encodings.reshape((batch_size, 4, 4, embedding_dim))
  return encodings

def remove_special_characters(word):
    return re.sub(r'^[^a-zA-Z0-9]+|[^a-zA-Z0-9]+$', '', word)

    
def extract_objs(text, width, height, unique_labels=False):
  """Returns objs for a string with "<loc>" and "<seg>" tokens."""
  objs = []
  seen = set()
  while text:
    m = _SEGMENT_DETECT_RE.match(text)
    if not m:
      break
    gs = list(m.groups())
    before = gs.pop(0)
    name = gs.pop()
    y1, x1, y2, x2 = [int(x) / 1024 for x in gs[:4]]
    
    y1, x1, y2, x2 = map(round, (y1*height, x1*width, y2*height, x2*width))
    seg_indices = gs[4:20]
    mask=None
    content = m.group()
    if before:
      objs.append(dict(content=before))
      content = content[len(before):]
    while unique_labels and name in seen:
      name = (name or '') + "'"
    seen.add(name)
    objs.append(dict(
        content=content, xyxy=(x1, y1, x2, y2), mask=mask, name=name))
    text = text[len(before) + len(content):]

  if text:
    objs.append(dict(content=text))

  return objs




with gr.Blocks() as demo:
  gr.Markdown("## Zero-shot Object Tracking with PaliGemma")
  gr.Markdown("This is a demo for zero-shot object tracking using [PaliGemma](https://huggingface.co/google/paligemma-3b-mix-448) vision language model by Google.")
  gr.Markdown("Simply upload a video and enter the candidate labels, or try the example below. Text input should be ; separated. 👇")
  with gr.Tab(label="Video"):
    with gr.Row():
        input_video = gr.Video(
            label='Input Video'
        )
        output_video = gr.Video(
            label='Output Video'
        )
    with gr.Row():
        candidate_labels = gr.Textbox(
            label='Labels',
            placeholder='Labels separated by a comma',
        )
        submit = gr.Button()
    gr.Examples(
        fn=process_video,
        examples=[["./cats.mp4", "bird ; cat"]],
        inputs=[
            input_video,
            candidate_labels,
            
        ],
        outputs=output_video
    )

  submit.click(
      fn=process_video,
      inputs=[input_video, candidate_labels],
      outputs=output_video
  )

demo.launch(debug=False, show_error=True)