Minor fixes

Author: Jimmy Wu

CLIP_GradCAM_Visualization.ipynb

@@ -44,23 +44,20 @@
     "!pip install ftfy regex tqdm matplotlib opencv-python scipy scikit-image\n",
     "!pip install git+https://github.com/openai/CLIP.git\n",
     "\n",
+    "import urllib.request\n",
     "import numpy as np\n",
     "import torch\n",
-    "import os\n",
-    "import torch.nn as nn\n",
     "import torch.nn.functional as F\n",
-    "import cv2\n",
-    "import urllib.request\n",
     "import matplotlib.pyplot as plt\n",
     "import clip\n",
     "from PIL import Image\n",
-    "from skimage import transform as skimage_transform\n",
-    "from scipy.ndimage import filters"
+    "from scipy.ndimage import filters\n",
+    "from torch import nn"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 34,
+   "execution_count": 2,
    "metadata": {
     "cellView": "form",
     "id": "caPbAhFlRBwT"
@@ -91,23 +88,23 @@
     "    return attn_map\n",
     "\n",
     "def viz_attn(img, attn_map, blur=True):\n",
-    "    fig, axes = plt.subplots(1, 2, figsize=(10, 5))\n",
+    "    _, axes = plt.subplots(1, 2, figsize=(10, 5))\n",
     "    axes[0].imshow(img)\n",
     "    axes[1].imshow(getAttMap(img, attn_map, blur))\n",
     "    for ax in axes:\n",
     "        ax.axis(\"off\")\n",
     "    plt.show()\n",
     "    \n",
     "def load_image(img_path, resize=None):\n",
-    "    image = Image.open(image_path).convert(\"RGB\")\n",
+    "    image = Image.open(img_path).convert(\"RGB\")\n",
     "    if resize is not None:\n",
     "        image = image.resize((resize, resize))\n",
     "    return np.asarray(image).astype(np.float32) / 255."
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 3,
    "metadata": {
     "cellView": "form",
     "id": "XziodsCqVC2A"
@@ -209,7 +206,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 45,
+   "execution_count": 4,
    "metadata": {
     "cellView": "form",
     "colab": {

CLIP_Patch_Detection.ipynb

@@ -42,14 +42,12 @@
     "!pip install ftfy regex tqdm matplotlib\n",
     "!pip install git+https://github.com/openai/CLIP.git\n",
     "\n",
-    "import numpy as np\n",
-    "import torch\n",
-    "import torch.nn as nn\n",
-    "import torch.nn.functional as F\n",
     "import math\n",
     "import urllib.request\n",
-    "import matplotlib.pyplot as plt\n",
     "import clip\n",
+    "import matplotlib.pyplot as plt\n",
+    "import numpy as np\n",
+    "import torch\n",
     "from PIL import Image\n",
     "from torchvision import transforms"
    ]
@@ -68,7 +66,7 @@
     "#@markdown Some helper functions for loading, patchifying and visualizing images.\n",
     "\n",
     "def load_image(img_path, resize=None, pil=False):\n",
-    "    image = Image.open(image_path).convert(\"RGB\")\n",
+    "    image = Image.open(img_path).convert(\"RGB\")\n",
     "    if resize is not None:\n",
     "        image = image.resize((resize, resize))\n",
     "    if pil:\n",
@@ -79,7 +77,7 @@
     "    # x: num_patches, 3, patch_size, patch_size\n",
     "    n = x.shape[0]\n",
     "    nrows = int(math.sqrt(n))\n",
-    "    fig, axes = plt.subplots(nrows, nrows, figsize=figsize)\n",
+    "    _, axes = plt.subplots(nrows, nrows, figsize=figsize)\n",
     "    for i, ax in enumerate(axes.flatten()):            \n",
     "        im = x[i].permute(1, 2, 0).numpy()\n",
     "        im = (im * 255.).round().astype(np.uint8)\n",

CLIP_Zero_shot_Detector.ipynb

@@ -42,19 +42,15 @@
     "!pip install ftfy regex tqdm matplotlib selectivesearch\n",
     "!pip install git+https://github.com/openai/CLIP.git\n",
     "\n",
-    "import numpy as np\n",
-    "import torch\n",
-    "import torch.nn as nn\n",
-    "import torch.nn.functional as F\n",
-    "import math\n",
     "import urllib.request\n",
+    "from collections import OrderedDict\n",
+    "import clip\n",
     "import matplotlib.pyplot as plt\n",
     "import matplotlib.patches as mpatches\n",
-    "import clip\n",
-    "from PIL import Image\n",
-    "from torchvision import transforms\n",
+    "import numpy as np\n",
     "import selectivesearch\n",
-    "from collections import OrderedDict"
+    "import torch\n",
+    "from PIL import Image"
    ]
   },
   {
@@ -71,7 +67,7 @@
     "#@markdown Some helper functions for loading, patchifying and visualizing images.\n",
     "\n",
     "def load_image(img_path, resize=None, pil=False):\n",
-    "    image = Image.open(image_path).convert(\"RGB\")\n",
+    "    image = Image.open(img_path).convert(\"RGB\")\n",
     "    if resize is not None:\n",
     "        image = image.resize((resize, resize))\n",
     "    if pil:\n",
@@ -118,7 +114,7 @@
     "\n",
     "    top_idx = dets[:, -1].argmax()\n",
     "\n",
-    "    fig, ax = plt.subplots(figsize=(12, 12))\n",
+    "    _, ax = plt.subplots(figsize=(12, 12))\n",
     "    ax.imshow(im, aspect='equal')\n",
     "    for i in inds:\n",
     "        bbox = dets[i, :4]\n",
@@ -231,7 +227,7 @@
     "candidates = candidates[:topk]\n",
     "\n",
     "# Display topk bounding boxes.\n",
-    "fig, ax = plt.subplots(ncols=1, nrows=1, figsize=(8, 8))\n",
+    "_, ax = plt.subplots(ncols=1, nrows=1, figsize=(8, 8))\n",
     "ax.imshow(img)\n",
     "for x, y, w, h in candidates:\n",
     "    rect = mpatches.Rectangle(\n",

CLIP_reCAPTCHA.ipynb

@@ -42,17 +42,15 @@
     "!pip install ftfy regex tqdm matplotlib bs4\n",
     "!pip install git+https://github.com/openai/CLIP.git\n",
     "\n",
+    "import urllib.request\n",
+    "import clip\n",
+    "import matplotlib.pyplot as plt\n",
     "import numpy as np\n",
     "import torch\n",
-    "import torch.nn as nn\n",
     "import torch.nn.functional as F\n",
-    "import math\n",
-    "import urllib.request\n",
-    "import matplotlib.pyplot as plt\n",
-    "import clip\n",
+    "from bs4 import BeautifulSoup\n",
     "from PIL import Image\n",
     "from torchvision import transforms\n",
-    "from bs4 import BeautifulSoup\n",
     "from torchvision.utils import make_grid"
    ]
   },
@@ -70,7 +68,7 @@
     "#@markdown Some helper functions for loading, patchifying and visualizing images.\n",
     "\n",
     "def load_image(img_path, resize=None, pil=False):\n",
-    "    image = Image.open(image_path).convert(\"RGB\")\n",
+    "    image = Image.open(img_path).convert(\"RGB\")\n",
     "    if resize is not None:\n",
     "        image = image.resize((resize, resize))\n",
     "    if pil:\n",
@@ -81,7 +79,6 @@
     "def viz_patches(x, figsize=None, topk=None, t=5, title=None):\n",
     "    color = (0, 255, 0)\n",
     "    n = x.shape[0]\n",
-    "    nrows = int(math.sqrt(n))\n",
     "    images = []\n",
     "    for i in range(n):\n",
     "        im = x[i].permute(1, 2, 0).numpy()\n",
@@ -160,7 +157,6 @@
     "url_contents = urllib.request.urlopen(URL).read()\n",
     "soup = BeautifulSoup(url_contents, \"html\")\n",
     "instruction = soup.find(\"div\", {\"class\": \"rc-imageselect-desc-no-canonical\"}).get_text()\n",
-    "caption = instruction.split(' ')[-1]\n",
     "image = soup.find(\"img\", {\"class\": \"fbc-imageselect-payload\"})\n",
     "image_url = f\"https://www.google.com/{image['src']}\"\n",
     "image_path = 'image.png'\n",
@@ -225,8 +221,7 @@
     "    patch_embs = patch_embs / patch_embs.norm(dim=-1, keepdim=True)\n",
     "    text_embs = text_embs / text_embs.norm(dim=-1, keepdim=True)\n",
     "    sim = patch_embs @ text_embs.t()\n",
-    "    idx_max = sim.argmax().item()\n",
-    "    topk_values, topk_idxs = torch.topk(sim.flatten(), topk)\n",
+    "    _, topk_idxs = torch.topk(sim.flatten(), topk)\n",
     "    topk_idxs = topk_idxs.cpu().numpy().tolist()\n",
     "\n",
     "viz_patches(\n",