initial commit of space using maskformer-satellite-trees model

.gitignore

@@ -0,0 +1,2 @@
+.DS_Store
+

README.md

@@ -1,13 +1,12 @@
----
-title: Maskformer Satellite Trees Gradio
-emoji: 👀
+
+title: Maskformer Satellite+Tree Gradio
+emoji: 🛰
 colorFrom: blue
-colorTo: yellow
+colorTo: indigo
 sdk: gradio
-sdk_version: 3.16.2
+sdk_version: 3.16.1
 app_file: app.py
+models:
+    - "thiagohersan/maskformer-satellite-trees"
 pinned: false
-license: other
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+license: cc-by-nc-sa-4.0

app.ipynb

@@ -0,0 +1,116 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import gradio as gr\n",
+    "import numpy as np\n",
+    "from PIL import Image as PImage\n",
+    "from torchvision import transforms as T\n",
+    "from transformers import MaskFormerForInstanceSegmentation, MaskFormerImageProcessor"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "ade_mean=[0.485, 0.456, 0.406]\n",
+    "ade_std=[0.229, 0.224, 0.225]\n",
+    "\n",
+    "model_id = f\"thiagohersan/maskformer-satellite-trees\""
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# preprocessor = MaskFormerImageProcessor.from_pretrained(model_id)\n",
+    "preprocessor = MaskFormerImageProcessor(\n",
+    "    do_resize=False,\n",
+    "    do_normalize=False,\n",
+    "    do_rescale=False,\n",
+    "    ignore_index=255,\n",
+    "    reduce_labels=False\n",
+    ")\n",
+    "\n",
+    "model = MaskFormerForInstanceSegmentation.from_pretrained(model_id)\n",
+    "\n",
+    "test_transform = T.Compose([\n",
+    "    T.ToTensor(),\n",
+    "    T.Normalize(mean=ade_mean, std=ade_std)\n",
+    "])\n",
+    "\n",
+    "with PImage.open(\"../color-filter-calculator/assets/Artshack_screen.jpg\") as img:\n",
+    "    img_size = (img.height, img.width)\n",
+    "    norm_image = test_transform(np.array(img))\n",
+    "    inputs = preprocessor(images=norm_image, return_tensors=\"pt\")\n",
+    "    "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "outputs = model(**inputs)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "results = preprocessor.post_process_semantic_segmentation(outputs=outputs, target_sizes=[img_size])[0]\n",
+    "results = results.numpy()\n",
+    "\n",
+    "labels = np.unique(results)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "for label_id in labels:\n",
+    "    print(model.config.id2label[label_id])"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3.8.15 ('hf-gradio')",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.8.15"
+  },
+  "orig_nbformat": 4,
+  "vscode": {
+   "interpreter": {
+    "hash": "4888b226c77b860705e4be316b14a092026f41c3585ee0ddb38f3008c0cb495e"
+   }
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

app.py

@@ -0,0 +1,77 @@
+import gradio as gr
+import numpy as np
+from torchvision import transforms as T
+from transformers import MaskFormerForInstanceSegmentation, MaskFormerImageProcessor
+
+
+ade_mean=[0.485, 0.456, 0.406]
+ade_std=[0.229, 0.224, 0.225]
+
+palette = [
+    [120, 120, 120], [4, 200, 4], [180, 120, 120], [6, 230, 230],
+    [80, 50, 50], [120, 120, 80], [140, 140, 140], [204, 5, 255]
+]
+
+model_id = f"thiagohersan/maskformer-satellite-trees"
+
+# preprocessor = MaskFormerImageProcessor.from_pretrained(model_id)
+preprocessor = MaskFormerImageProcessor(
+    do_resize=False,
+    do_normalize=False,
+    do_rescale=False,
+    ignore_index=255,
+    reduce_labels=False
+)
+
+model = MaskFormerForInstanceSegmentation.from_pretrained(model_id)
+
+test_transform = T.Compose([
+    T.ToTensor(),
+    T.Normalize(mean=ade_mean, std=ade_std)
+])
+
+def visualize_instance_seg_mask(img_in, mask, id2label):
+    img_out = np.zeros((mask.shape[0], mask.shape[1], 3))
+    image_total_pixels = mask.shape[0] * mask.shape[1]
+    label_ids = np.unique(mask)
+
+    id2color = {id: palette[id] for id in label_ids}
+    id2count = {id: 0 for id in label_ids}
+
+    for i in range(img_out.shape[0]):
+      for j in range(img_out.shape[1]):
+        img_out[i, j, :] = id2color[mask[i, j]]
+        id2count[mask[i, j]] = id2count[mask[i, j]] + 1
+
+    image_res = (0.5 * img_in + 0.5 * img_out) / 255
+
+    dataframe = [[
+        f"{id2label[id]}",
+        f"{(100 * id2count[id] / image_total_pixels):.2f} %",
+        f"{np.sqrt(id2count[id] / image_total_pixels):.2f} m"
+        ] for id in label_ids if 'tree' in id2label[id]]
+
+    return image_res, dataframe
+
+
+def query_image(img):
+    img_size = (img.shape[0], img.shape[1])
+    inputs = preprocessor(images=test_transform(np.array(img)), return_tensors="pt")
+
+    outputs = model(**inputs)
+
+    results = preprocessor.post_process_semantic_segmentation(outputs=outputs, target_sizes=[img_size])[0]
+    results = visualize_instance_seg_mask(img, results.numpy(), model.config.id2label)
+    return results
+
+
+demo = gr.Interface(
+    query_image,
+    inputs=[gr.Image(label="Input Image")],
+    outputs=[gr.Image(label="Trees"), gr.DataFrame(headers=None, label="Area Info")],
+    title="maskformer-satellite-trees",
+    allow_flagging="never",
+    analytics_enabled=None
+)
+
+demo.launch(show_api=False)

requirements.txt

@@ -0,0 +1,4 @@
+scipy
+torch
+torchvision
+transformers