first commit

app.py

@@ -1,37 +1,12 @@
-"""
-CellVision AI - Intelligent Cell Imaging Analysis
-
-This module provides a Gradio web application for performing intelligent cell imaging analysis
-using the PaliGemma model from Google. The app allows users to segment or detect cells in images
-and generate descriptive text based on the input image and prompt.
-
-Dependencies:
-- gradio
-- transformers
-- torch
-- jax
-- flax
-- spaces
-- PIL
-- numpy
-- huggingface_hub
-
-"""
-
-import os
-import functools
-import re
-
-import PIL.Image
 import gradio as gr
-import numpy as np
-import torch
-import jax
-import jax.numpy as jnp
-import flax.linen as nn
+from typing import *
+from pillow_heif import register_heif_opener
+register_heif_opener()
+import vision_agent as va
+from vision_agent.tools import register_tool
+
+from vision_agent.tools import load_image, owl_v2, overlay_bounding_boxes, save_image
 
-from transformers import PaliGemmaForConditionalGeneration, PaliGemmaProcessor
-from peft import PeftConfig, PeftModel
 from huggingface_hub import login
 import spaces
 
@@ -39,278 +14,38 @@ import spaces
 hf_token = os.getenv("HF_TOKEN")
 login(token=hf_token, add_to_git_credential=True)
 
-
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-
-model_id = "google/paligemma-3b-pt-224"
-adapter_model_id = "dwb2023/paligemma-tumor-detection-ft"
-
-# model_id = "google/paligemma-3b-ft-refcoco-seg-224"
-model = PaliGemmaForConditionalGeneration.from_pretrained(model_id).eval().to(device)
-
-model = PeftModel.from_pretrained(model, adapter_model_id).to(device)
-model = model.merge_and_unload()
-model.save_pretrained("merged_adapters")
-
-processor = PaliGemmaProcessor.from_pretrained(model_id)
-
-@spaces.GPU(duration=120)
-def infer(
-    image: PIL.Image.Image,
-    text: str,
-    max_new_tokens: int
-) -> str:
-    """
-    Perform inference using the PaliGemma model.
-
-    Args:
-        image (PIL.Image.Image): Input image.
-        text (str): Input text prompt.
-        max_new_tokens (int): Maximum number of new tokens to generate.
-
-    Returns:
-        str: Generated text based on the input image and prompt.
-    """
-    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_segmentation(input_image, input_text):
-    """
-    Parse segmentation output tokens into masks and bounding boxes.
-
-    Args:
-        input_image (PIL.Image.Image): Input image.
-        input_text (str): Input text specifying entities to segment or detect.
-
-    Returns:
-        tuple: A tuple containing the annotated image and a boolean indicating if annotations are present.
-    """
-    out = infer(input_image, input_text, max_new_tokens=100)
-    objs = extract_objs(out.lstrip("\n"), input_image.size[0], input_image.size[1], unique_labels=True)
-    labels = set(obj.get('name') for obj in objs if obj.get('name'))
-    color_map = {l: COLORS[i % len(COLORS)] for i, l in enumerate(labels)}
-    highlighted_text = [(obj['content'], obj.get('name')) for obj in objs]
-    annotated_img = (
-        input_image,
-        [
-            (
-                obj['mask'] if obj.get('mask') is not None else obj['xyxy'],
-                obj['name'] or '',
-            )
-            for obj in objs
-            if 'mask' in obj or 'xyxy' in obj
-        ],
-    )
-    has_annotations = bool(annotated_img[1])
-    return annotated_img
-
-
-### Postprocessing Utils for Segmentation Tokens
-
-_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*([^;<>]+)? ?(?:; )?',
-)
-
-COLORS = ['#4285f4', '#db4437', '#f4b400', '#0f9d58', '#e48ef1']
-
-def _get_params(checkpoint):
-    """
-    Convert PyTorch checkpoint to Flax params.
-
-    Args:
-        checkpoint (dict): PyTorch checkpoint dictionary.
-
-    Returns:
-        dict: Flax parameters.
-    """
-    def transp(kernel):
-        return np.transpose(kernel, (2, 3, 1, 0))
-
-    def conv(name):
-        return {
-            'bias': checkpoint[name + '.bias'],
-            'kernel': transp(checkpoint[name + '.weight']),
-        }
-
-    def resblock(name):
-        return {
-            'Conv_0': conv(name + '.0'),
-            'Conv_1': conv(name + '.2'),
-            'Conv_2': conv(name + '.4'),
-        }
-
-    return {
-        '_embeddings': checkpoint['_vq_vae._embedding'],
-        'Conv_0': conv('decoder.0'),
-        'ResBlock_0': resblock('decoder.2.net'),
-        'ResBlock_1': resblock('decoder.3.net'),
-        'ConvTranspose_0': conv('decoder.4'),
-        'ConvTranspose_1': conv('decoder.6'),
-        'ConvTranspose_2': conv('decoder.8'),
-        'ConvTranspose_3': conv('decoder.10'),
-        'Conv_1': conv('decoder.12'),
-    }
-
-
-def _quantized_values_from_codebook_indices(codebook_indices, embeddings):
-    """
-    Get quantized values from codebook indices.
-
-    Args:
-        codebook_indices (jax.numpy.ndarray): Codebook indices.
-        embeddings (jax.numpy.ndarray): Embeddings.
-
-    Returns:
-        jax.numpy.ndarray: Quantized values.
-    """
-    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
-
-
-@functools.cache
-def _get_reconstruct_masks():
-    """
-    Reconstruct masks from codebook indices.
-
-    Returns:
-        function: A function that expects indices shaped `[B, 16]` of dtype int32, each
-        ranging from 0 to 127 (inclusive), and returns decoded masks sized
-        `[B, 64, 64, 1]`, of dtype float32, in range [-1, 1].
-    """
-
-    class ResBlock(nn.Module):
-        features: int
-
-        @nn.compact
-        def __call__(self, x):
-            original_x = x
-            x = nn.Conv(features=self.features, kernel_size=(3, 3), padding=1)(x)
-            x = nn.relu(x)
-            x = nn.Conv(features=self.features, kernel_size=(3, 3), padding=1)(x)
-            x = nn.relu(x)
-            x = nn.Conv(features=self.features, kernel_size=(1, 1), padding=0)(x)
-            return x + original_x
-
-    class Decoder(nn.Module):
-        """Upscales quantized vectors to mask."""
-
-        @nn.compact
-        def __call__(self, x):
-            num_res_blocks = 2
-            dim = 128
-            num_upsample_layers = 4
-
-            x = nn.Conv(features=dim, kernel_size=(1, 1), padding=0)(x)
-            x = nn.relu(x)
-
-            for _ in range(num_res_blocks):
-                x = ResBlock(features=dim)(x)
-
-            for _ in range(num_upsample_layers):
-                x = nn.ConvTranspose(
-                    features=dim,
-                    kernel_size=(4, 4),
-                    strides=(2, 2),
-                    padding=2,
-                    transpose_kernel=True,
-                )(x)
-                x = nn.relu(x)
-                dim //= 2
-
-            x = nn.Conv(features=1, kernel_size=(1, 1), padding=0)(x)
-
-            return x
-
-    def reconstruct_masks(codebook_indices):
-        """
-        Reconstruct masks from codebook indices.
-
-        Args:
-            codebook_indices (jax.numpy.ndarray): Codebook indices.
-
-        Returns:
-            jax.numpy.ndarray: Reconstructed masks.
-        """
-        quantized = _quantized_values_from_codebook_indices(
-            codebook_indices, params['_embeddings']
-        )
-        return Decoder().apply({'params': params}, quantized)
-
-    with open(_MODEL_PATH, 'rb') as f:
-        params = _get_params(dict(np.load(f)))
-
-    return jax.jit(reconstruct_masks, backend='cpu')
-
-def extract_objs(text, width, height, unique_labels=False):
+def detect_brain_tumor(image_path: str, output_path: str, debug: bool = False) -> None:
     """
-    Extract objects from text containing "<loc>" and "<seg>" tokens.
-
-    Args:
-        text (str): Input text containing "<loc>" and "<seg>" tokens.
-        width (int): Width of the image.
-        height (int): Height of the image.
-        unique_labels (bool, optional): Whether to enforce unique labels. Defaults to False.
+    Detects a brain tumor in the given image and saves the image with bounding boxes.
 
-    Returns:
-        list: List of extracted objects.
+    Parameters:
+        image_path (str): The path to the input image.
+        output_path (str): The path to save the output image with bounding boxes.
+        debug (bool): Flag to enable logging for debugging purposes.
     """
-    objs = []
-    seen = set()
-    while text:
-        m = _SEGMENT_DETECT_RE.match(text)
-        if not m:
-            break
-        print("m", m)
-        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]
-        if seg_indices[0] is None:
-            mask = None
-        else:
-            seg_indices = np.array([int(x) for x in seg_indices], dtype=np.int32)
-            m64, = _get_reconstruct_masks()(seg_indices[None])[..., 0]
-            m64 = np.clip(np.array(m64) * 0.5 + 0.5, 0, 1)
-            m64 = PIL.Image.fromarray((m64 * 255).astype('uint8'))
-            mask = np.zeros([height, width])
-            if y2 > y1 and x2 > x1:
-                mask[y1:y2, x1:x2] = np.array(m64.resize([x2 - x1, y2 - y1])) / 255.0
-
-        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
+    # Step 1: Load the image
+    image = load_image(image_path)
+    if debug:
+        print(f"Image loaded from {image_path}")
+
+    # Step 2: Detect brain tumor using owl_v2
+    prompt = "detect brain tumor"
+    detections = owl_v2(prompt, image)
+    if debug:
+        print(f"Detections: {detections}")
+
+    # Step 3: Overlay bounding boxes on the image
+    image_with_bboxes = overlay_bounding_boxes(image, detections)
+    if debug:
+        print("Bounding boxes overlaid on the image")
+
+    # Step 4: Save the resulting image
+    save_image(image_with_bboxes, output_path)
+    if debug:
+        print(f"Image saved to {output_path}")
+
+# Example usage (uncomment to run):
+# detect_brain_tumor("/content/drive/MyDrive/kaggle/datasets/brain-tumor-image-dataset-semantic-segmentation_old/train_categories/1385_jpg.rf.3c67cb92e2922dba0e6dba86f69df40b.jpg", "/content/drive/MyDrive/kaggle/datasets/brain-tumor-image-dataset-semantic-segmentation_old/output/1385_jpg.rf.3c67cb92e2922dba0e6dba86f69df40b.jpg", debug=True)
 
 #########
 
@@ -319,90 +54,39 @@ IMAGE_PROMPT="Are these cells healthy or cancerous?"
 
 with gr.Blocks(css="style.css") as demo:
     gr.Markdown(INTRO_TEXT)
-    with gr.Tab("Segment/Detect"):
-        with gr.Row():
-            with gr.Column():
-                image = gr.Image(type="pil")
-                seg_input = gr.Text(label="Entities to Segment/Detect")
-        
-            with gr.Column():
-                annotated_image = gr.AnnotatedImage(label="Output")
-
-        seg_btn = gr.Button("Submit")    
-        examples = [["./examples/cnmc1.bmp", "segment cancerous cells"],
-                    ["./examples/cnmc2.bmp", "detect cancerous cells"],
-                    ["./examples/cnmc3.bmp", "segment healthy cells"],
-                    ["./examples/cnmc4.bmp", "detect healthy cells"],
-                    ["./examples/cnmc5.bmp", "segment cancerous cells"],
-                    ["./examples/cnmc6.bmp", "detect cancerous cells"],
-                    ["./examples/cnmc7.bmp", "segment healthy cells"],
-                    ["./examples/cnmc8.bmp", "detect healthy cells"],
-                    ["./examples/cnmc9.bmp", "segment cancerous cells"],
-                    ["./examples/cart1.jpg", "segment cells"],
-                    ["./examples/cart1.jpg", "detect cells"],
-                    ["./examples/cart2.jpg", "segment cells"],
-                    ["./examples/cart2.jpg", "detect cells"],
-                    ["./examples/cart3.jpg", "segment cells"],
-                    ["./examples/cart3.jpg", "detect cells"]]
-        gr.Examples(
-            examples=examples,
-            inputs=[image, seg_input],
-        )
-        seg_inputs = [
-            image,
-            seg_input
-            ]
-        seg_outputs = [
-            annotated_image
-        ]
-        seg_btn.click(
-            fn=parse_segmentation,
-            inputs=seg_inputs,
-            outputs=seg_outputs,
-        )
-    with gr.Tab("Text Generation"):
+    with gr.Tab("Agentic Detection"):
         with gr.Row():
             with gr.Column():
                 image = gr.Image(type="pil")
             with gr.Column():
                 text_input = gr.Text(label="Input Text")
                 text_output = gr.Text(label="Text Output")
-                tokens = gr.Slider(
-                    label="Max New Tokens",
-                    info="Set to larger for longer generation.",
-                    minimum=10,
-                    maximum=100,
-                    value=50,
-                    step=10,
-                )
         chat_btn = gr.Button()
 
         chat_inputs = [
             image,
-            text_input,
-            tokens
+            "./output/tmp.jpg",
             ]
         chat_outputs = [
             text_output
         ]
         chat_btn.click(
-            fn=infer,
+            fn=detect_brain_tumor,
             inputs=chat_inputs,
             outputs=chat_outputs,
         )
         
-        examples = [["./examples/cnmc1.bmp", IMAGE_PROMPT],
-                    ["./examples/cnmc2.bmp", IMAGE_PROMPT],
-                    ["./examples/cnmc3.bmp", IMAGE_PROMPT],
-                    ["./examples/cnmc4.bmp", IMAGE_PROMPT],
-                    ["./examples/cnmc5.bmp", IMAGE_PROMPT],
-                    ["./examples/cnmc6.bmp", IMAGE_PROMPT],
-                    ["./examples/cnmc7.bmp", IMAGE_PROMPT],
-                    ["./examples/cnmc8.bmp", IMAGE_PROMPT],
-                    ["./examples/cnmc9.bmp", IMAGE_PROMPT],
-                    ["./examples/cart1.jpg", IMAGE_PROMPT],
-                    ["./examples/cart2.jpg", IMAGE_PROMPT],
-                    ["./examples/cart3.jpg", IMAGE_PROMPT]]
+        examples = [["./examples/194_jpg.rf.3e3dd592d034bb5ee27a978553819f42.jpg"],
+                    ["./examples/239_jpg.rf.3dcc0799277fb78a2ab21db7761ccaeb.jpg"],
+                    ["./examples/1385_jpg.rf.3c67cb92e2922dba0e6dba86f69df40b.jpg"],
+                    ["./examples/1491_jpg.rf.3c658e83538de0fa5a3f4e13d7d85f12.jpg"],
+                    ["./examples/1550_jpg.rf.3d067be9580ec32dbee5a89c675d8459.jpg"],
+                    ["./examples/2256_jpg.rf.3afd7903eaf3f3c5aa8da4bbb928bc19.jpg"],
+                    ["./examples/1550_jpg.rf.3d067be9580ec32dbee5a89c675d8459.jpg"],
+                    ["./examples/1550_jpg.rf.3d067be9580ec32dbee5a89c675d8459.jpg"],
+                    ["./examples/2871_jpg.rf.3b6eadfbb369abc2b3bcb52b406b74f2.jpg"],
+                    ["./examples/2921_jpg.rf.3b952f91f27a6248091e7601c22323ad.jpg"],
+                    ]
         gr.Examples(
             examples=examples,
             inputs=chat_inputs,

examples/_annotations.coco.json

@@ -0,0 +1,388 @@
+{
+    "info": {
+        "year": "2023",
+        "version": "1",
+        "description": "Exported from roboflow.com",
+        "contributor": "",
+        "url": "https://public.roboflow.com/object-detection/undefined",
+        "date_created": "2023-08-19T04:37:54+00:00"
+    },
+    "licenses": [
+        {
+            "id": 1,
+            "url": "https://creativecommons.org/licenses/by/4.0/",
+            "name": "CC BY 4.0"
+        }
+    ],
+    "categories": [
+        {
+            "id": 0,
+            "name": "Tumor",
+            "supercategory": "none"
+        },
+        {
+            "id": 1,
+            "name": "0",
+            "supercategory": "Tumor"
+        },
+        {
+            "id": 2,
+            "name": "1",
+            "supercategory": "Tumor"
+        }
+    ],
+    "images": [
+        {
+            "id": 0,
+            "license": 1,
+            "file_name": "2256_jpg.rf.3afd7903eaf3f3c5aa8da4bbb928bc19.jpg",
+            "height": 640,
+            "width": 640,
+            "date_captured": "2023-08-19T04:37:54+00:00"
+        },
+        {
+            "id": 1,
+            "license": 1,
+            "file_name": "2871_jpg.rf.3b6eadfbb369abc2b3bcb52b406b74f2.jpg",
+            "height": 640,
+            "width": 640,
+            "date_captured": "2023-08-19T04:37:54+00:00"
+        },
+        {
+            "id": 2,
+            "license": 1,
+            "file_name": "2921_jpg.rf.3b952f91f27a6248091e7601c22323ad.jpg",
+            "height": 640,
+            "width": 640,
+            "date_captured": "2023-08-19T04:37:54+00:00"
+        },
+        {
+            "id": 3,
+            "license": 1,
+            "file_name": "2777_jpg.rf.3b60ea7f7e70552e70e41528052018bd.jpg",
+            "height": 640,
+            "width": 640,
+            "date_captured": "2023-08-19T04:37:54+00:00"
+        },
+        {
+            "id": 4,
+            "license": 1,
+            "file_name": "2860_jpg.rf.3bb87fa4f938af5abfb1e17676ec1dad.jpg",
+            "height": 640,
+            "width": 640,
+            "date_captured": "2023-08-19T04:37:54+00:00"
+        },
+        {
+            "id": 7,
+            "license": 1,
+            "file_name": "1491_jpg.rf.3c658e83538de0fa5a3f4e13d7d85f12.jpg",
+            "height": 640,
+            "width": 640,
+            "date_captured": "2023-08-19T04:37:54+00:00"
+        },
+        {
+            "id": 8,
+            "license": 1,
+            "file_name": "1385_jpg.rf.3c67cb92e2922dba0e6dba86f69df40b.jpg",
+            "height": 640,
+            "width": 640,
+            "date_captured": "2023-08-19T04:37:54+00:00"
+        },
+        {
+            "id": 11,
+            "license": 1,
+            "file_name": "1550_jpg.rf.3d067be9580ec32dbee5a89c675d8459.jpg",
+            "height": 640,
+            "width": 640,
+            "date_captured": "2023-08-19T04:37:54+00:00"
+        },
+        {
+            "id": 16,
+            "license": 1,
+            "file_name": "239_jpg.rf.3dcc0799277fb78a2ab21db7761ccaeb.jpg",
+            "height": 640,
+            "width": 640,
+            "date_captured": "2023-08-19T04:37:54+00:00"
+        },
+        {
+            "id": 17,
+            "license": 1,
+            "file_name": "194_jpg.rf.3e3dd592d034bb5ee27a978553819f42.jpg",
+            "height": 640,
+            "width": 640,
+            "date_captured": "2023-08-19T04:37:54+00:00"
+        }        
+    ],
+    "annotations": [
+        {
+            "id": 0,
+            "image_id": 0,
+            "category_id": 1,
+            "bbox": [
+                145,
+                239,
+                168.75,
+                162.5
+            ],
+            "area": 27421.875,
+            "segmentation": [
+                [
+                    313.75,
+                    238.75,
+                    145,
+                    238.75,
+                    145,
+                    401.25,
+                    313.75,
+                    401.25,
+                    313.75,
+                    238.75
+                ]
+            ],
+            "iscrowd": 0
+        },
+        {
+            "id": 1,
+            "image_id": 1,
+            "category_id": 1,
+            "bbox": [
+                194,
+                176,
+                148.75,
+                233.75
+            ],
+            "area": 34770.313,
+            "segmentation": [
+                [
+                    342.5,
+                    176.25,
+                    193.75,
+                    176.25,
+                    193.75,
+                    410,
+                    342.5,
+                    410,
+                    342.5,
+                    176.25
+                ]
+            ],
+            "iscrowd": 0
+        },
+        {
+            "id": 2,
+            "image_id": 2,
+            "category_id": 1,
+            "bbox": [
+                133,
+                173,
+                162.5,
+                185
+            ],
+            "area": 30062.5,
+            "segmentation": [
+                [
+                    295,
+                    172.5,
+                    132.5,
+                    172.5,
+                    132.5,
+                    357.5,
+                    295,
+                    357.5,
+                    295,
+                    172.5
+                ]
+            ],
+            "iscrowd": 0
+        },
+        {
+            "id": 3,
+            "image_id": 3,
+            "category_id": 1,
+            "bbox": [
+                245,
+                358,
+                138.75,
+                166.25
+            ],
+            "area": 23067.188,
+            "segmentation": [
+                [
+                    383.75,
+                    357.5,
+                    245,
+                    357.5,
+                    245,
+                    523.75,
+                    383.75,
+                    523.75,
+                    383.75,
+                    357.5
+                ]
+            ],
+            "iscrowd": 0
+        },
+        {
+            "id": 4,
+            "image_id": 4,
+            "category_id": 1,
+            "bbox": [
+                80,
+                189,
+                112.5,
+                132.5
+            ],
+            "area": 14906.25,
+            "segmentation": [
+                [
+                    192.5,
+                    188.75,
+                    80,
+                    188.75,
+                    80,
+                    321.25,
+                    192.5,
+                    321.25,
+                    192.5,
+                    188.75
+                ]
+            ],
+            "iscrowd": 0
+        },
+        {
+            "id": 7,
+            "image_id": 7,
+            "category_id": 2,
+            "bbox": [
+                350,
+                288,
+                42.5,
+                52.5
+            ],
+            "area": 2231.25,
+            "segmentation": [
+                [
+                    392.5,
+                    287.5,
+                    350,
+                    287.5,
+                    350,
+                    340,
+                    392.5,
+                    340,
+                    392.5,
+                    287.5
+                ]
+            ],
+            "iscrowd": 0
+        },
+        {
+            "id": 8,
+            "image_id": 8,
+            "category_id": 2,
+            "bbox": [
+                239,
+                250,
+                61.25,
+                87.5
+            ],
+            "area": 5359.375,
+            "segmentation": [
+                [
+                    300,
+                    250,
+                    238.75,
+                    250,
+                    238.75,
+                    337.5,
+                    300,
+                    337.5,
+                    300,
+                    250
+                ]
+            ],
+            "iscrowd": 0
+        },
+        {
+            "id": 11,
+            "image_id": 11,
+            "category_id": 2,
+            "bbox": [
+                283,
+                270,
+                111.25,
+                102.5
+            ],
+            "area": 11403.125,
+            "segmentation": [
+                [
+                    393.75,
+                    270,
+                    282.5,
+                    270,
+                    282.5,
+                    372.5,
+                    393.75,
+                    372.5,
+                    393.75,
+                    270
+                ]
+            ],
+            "iscrowd": 0
+        },
+        {
+            "id": 16,
+            "image_id": 16,
+            "category_id": 2,
+            "bbox": [
+                279,
+                236,
+                78.75,
+                76.25
+            ],
+            "area": 6004.688,
+            "segmentation": [
+                [
+                    357.5,
+                    236.25,
+                    278.75,
+                    236.25,
+                    278.75,
+                    312.5,
+                    357.5,
+                    312.5,
+                    357.5,
+                    236.25
+                ]
+            ],
+            "iscrowd": 0
+        },
+        {
+            "id": 17,
+            "image_id": 17,
+            "category_id": 2,
+            "bbox": [
+                286,
+                160,
+                188.75,
+                166.25
+            ],
+            "area": 31379.688,
+            "segmentation": [
+                [
+                    475,
+                    160,
+                    286.25,
+                    160,
+                    286.25,
+                    326.25,
+                    475,
+                    326.25,
+                    475,
+                    160
+                ]
+            ],
+            "iscrowd": 0
+        }        
+    ]
+}

examples/tumor1.json

@@ -0,0 +1,7 @@
+{
+  "name": "tumor1",
+  "comment": "",
+  "model": "",
+  "prompt": "detect cell tumor",
+  "license": ""
+}

examples/tumor10.json

@@ -0,0 +1,7 @@
+{
+  "name": "tumor10",
+  "comment": "",
+  "model": "",
+  "prompt": "detect cell tumor",
+  "license": ""
+}

examples/tumor2.json

@@ -0,0 +1,7 @@
+{
+  "name": "tumor2",
+  "comment": "",
+  "model": "",
+  "prompt": "detect cell tumor",
+  "license": ""
+}

examples/tumor3.json

@@ -0,0 +1,7 @@
+{
+  "name": "tumor3",
+  "comment": "",
+  "model": "",
+  "prompt": "detect cell tumor",
+  "license": ""
+}

examples/tumor4.json

@@ -0,0 +1,7 @@
+{
+  "name": "tumor4",
+  "comment": "",
+  "model": "",
+  "prompt": "detect cell tumor",
+  "license": ""
+}

examples/tumor5.json

@@ -0,0 +1,7 @@
+{
+  "name": "tumor5",
+  "comment": "",
+  "model": "",
+  "prompt": "detect cell tumor",
+  "license": ""
+}

examples/tumor6.json

@@ -0,0 +1,7 @@
+{
+  "name": "tumor6",
+  "comment": "",
+  "model": "",
+  "prompt": "detect cell tumor",
+  "license": ""
+}

examples/tumor7.json

@@ -0,0 +1,7 @@
+{
+  "name": "tumor7",
+  "comment": "",
+  "model": "",
+  "prompt": "detect cell tumor",
+  "license": ""
+}

examples/tumor8.json

@@ -0,0 +1,7 @@
+{
+  "name": "tumor8",
+  "comment": "",
+  "model": "",
+  "prompt": "detect cell tumor",
+  "license": ""
+}

examples/tumor9.json

@@ -0,0 +1,7 @@
+{
+  "name": "tumor9",
+  "comment": "",
+  "model": "",
+  "prompt": "detect cell tumor",
+  "license": ""
+}

requirements.txt

@@ -1,6 +1,2 @@
-git+https://github.com/huggingface/transformers.git
-torch
-jax
-flax
 spaces
-peft
+vision-agent

vae-oid.npz

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:5586010257b8536dddefab65e7755077f21d5672d5674dacf911f73ae95a4447
-size 8479556