Update tutorial dependencies and conversion notebooks

conversion/nvidia_tao_conversion.ipynb

@@ -99,7 +99,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -145,9 +145,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "%pip install -q pillow onnx onnxruntime modelconv==0.4.5 -U\n",
-    "%pip install -q depthai==3.0.0 -U\n",
-    "%pip install -q depthai-nodes==0.3.0 -U"
+    "%pip install -q pillow luxonis-ml==0.8.6 onnx==1.21.0 onnxruntime==1.23.2 modelconv==0.5.5 depthai==3.7.1 depthai-nodes==0.5.1\n",
+    "%pip install -q numpy==2.0.2"
    ]
   },
   {
@@ -243,10 +242,10 @@
     }
    ],
    "source": [
+    "import os\n",
     "import onnxruntime as ort\n",
     "import numpy as np\n",
     "import cv2\n",
-    "from scipy.spatial.distance import cosine\n",
     "import matplotlib.pyplot as plt\n",
     "\n",
     "# Load ONNX model\n",
@@ -285,8 +284,15 @@
     "    embeddings[img_path] = normalize_embedding(output.flatten())  # Flatten to 1D vector\n",
     "\n",
     "# Compute cosine similarity between pairs\n",
-    "def cosine_similarity(emb1, emb2):\n",
-    "    return 1 - cosine(emb1, emb2)  # Higher value = more similar\n",
+    "def cosine_similarity(emb1, emb2, eps=1e-12):\n",
+    "    emb1 = np.asarray(emb1, dtype=np.float32).ravel()\n",
+    "    emb2 = np.asarray(emb2, dtype=np.float32).ravel()\n",
+    "\n",
+    "    denom = np.linalg.norm(emb1) * np.linalg.norm(emb2)\n",
+    "    if denom < eps:\n",
+    "        return 0.0\n",
+    "\n",
+    "    return float(np.dot(emb1, emb2) / denom)\n",
     "\n",
     "pairs = [\n",
     "    (\"person_0_0.png\", \"person_0_1.png\"),  # Same person\n",
@@ -442,7 +448,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "%pip install -q hubai-sdk"
+    "%pip install -q -U hubai-sdk"
    ]
   },
   {
@@ -458,7 +464,13 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "HUBAI_API_KEY = \"<YOUR_HUBAI_API_KEY>\""
+    "import os\n",
+    "import getpass\n",
+    "\n",
+    "if not os.environ.get(\"HUBAI_API_KEY\"):\n",
+    "    os.environ[\"HUBAI_API_KEY\"] = getpass.getpass(\"Enter your HubAI API key: \")\n",
+    "\n",
+    "print(\"HUBAI_API_KEY is set:\", bool(os.environ.get(\"HUBAI_API_KEY\")))"
    ]
   },
   {
@@ -467,19 +479,25 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "from pathlib import Path\n",
     "from hubai_sdk import HubAIClient\n",
     "\n",
-    "client = HubAIClient(api_key=HUBAI_API_KEY)\n",
+    "client = HubAIClient(api_key=os.environ[\"HUBAI_API_KEY\"])\n",
+    "\n",
+    "ARTIFACT_DIR = Path(\"reidentification-nvidia-tao-exported-to-rvc4\")\n",
+    "ARTIFACT_DIR.mkdir(parents=True, exist_ok=True)\n",
     "\n",
     "response = client.convert.RVC4(\n",
     "    path=\"resnet50_market1501_aicity156.tar.xz\",\n",
     "    name=\"ReIdentification NVIDIA TAO\",\n",
     "    description_short=\"ReIdentificationNet takes cropped images of a person from different perspectives as network input and outputs the embedding features for that person.\",\n",
     "    license_type=\"MIT\",\n",
     "    is_public=False,\n",
+    "    output_dir=str(ARTIFACT_DIR),\n",
     ")\n",
     "\n",
-    "model_path = response.downloaded_path\n",
+    "MODEL_PATH = Path(response.downloaded_path)\n",
+    "print(\"Model artifact:\", MODEL_PATH)\n",
     "\n",
     "# =============================================================================\n",
     "# RVC4 conversion\n",