train

app.ipynb

@@ -70,9 +70,7 @@
    "cell_type": "code",
    "execution_count": 5,
    "id": "6e0bf9da",
-   "metadata": {
-    "scrolled": false
-   },
+   "metadata": {},
    "outputs": [
     {
      "data": {
@@ -322,7 +320,10 @@
    "cell_type": "code",
    "execution_count": 16,
    "id": "82774c08",
-   "metadata": {},
+   "metadata": {
+    "scrolled": true,
+    "tags": []
+   },
    "outputs": [
     {
      "data": {
@@ -791,7 +792,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.5"
+   "version": "3.7.11"
   },
   "toc": {
    "base_numbering": 1,

train.ipynb

@@ -10,7 +10,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": 1,
    "id": "44eb0ad3",
    "metadata": {},
    "outputs": [],
@@ -21,7 +21,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": 7,
    "id": "d838c0b3",
    "metadata": {},
    "outputs": [],
@@ -145,19 +145,62 @@
     "learn.fine_tune(3)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "477ef53a-4b5c-4a07-81c2-95b8e7397cac",
+   "metadata": {},
+   "source": [
+    "We could try a better model, based on [this analysis](https://www.kaggle.com/code/jhoward/which-image-models-are-best/). The convnext models work great!"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": 6,
-   "id": "931ab602",
+   "id": "6ee4197a-25be-48ac-a167-903bec5186b1",
    "metadata": {},
    "outputs": [
     {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "Downloading: \"https://dl.fbaipublicfiles.com/convnext/convnext_tiny_22k_224.pth\" to /root/.cache/torch/hub/checkpoints/convnext_tiny_22k_224.pth\n"
-     ]
-    },
+     "data": {
+      "text/plain": [
+       "['convnext_base',\n",
+       " 'convnext_base_384_in22ft1k',\n",
+       " 'convnext_base_in22ft1k',\n",
+       " 'convnext_base_in22k',\n",
+       " 'convnext_large',\n",
+       " 'convnext_large_384_in22ft1k',\n",
+       " 'convnext_large_in22ft1k',\n",
+       " 'convnext_large_in22k',\n",
+       " 'convnext_nano_hnf',\n",
+       " 'convnext_small',\n",
+       " 'convnext_small_384_in22ft1k',\n",
+       " 'convnext_small_in22ft1k',\n",
+       " 'convnext_small_in22k',\n",
+       " 'convnext_tiny',\n",
+       " 'convnext_tiny_384_in22ft1k',\n",
+       " 'convnext_tiny_hnf',\n",
+       " 'convnext_tiny_hnfd',\n",
+       " 'convnext_tiny_in22ft1k',\n",
+       " 'convnext_tiny_in22k',\n",
+       " 'convnext_xlarge_384_in22ft1k',\n",
+       " 'convnext_xlarge_in22ft1k',\n",
+       " 'convnext_xlarge_in22k']"
+      ]
+     },
+     "execution_count": 6,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "timm.list_models('convnext*')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "67d34f88-a580-48b9-9b42-e9d0c7e3e870",
+   "metadata": {},
+   "outputs": [
     {
      "data": {
       "text/html": [
@@ -174,9 +217,9 @@
        "  <tbody>\n",
        "    <tr>\n",
        "      <td>0</td>\n",
-       "      <td>1.121368</td>\n",
-       "      <td>0.267543</td>\n",
-       "      <td>0.080514</td>\n",
+       "      <td>1.113290</td>\n",
+       "      <td>0.253742</td>\n",
+       "      <td>0.087957</td>\n",
        "      <td>00:24</td>\n",
        "    </tr>\n",
        "  </tbody>\n",
@@ -205,23 +248,23 @@
        "  <tbody>\n",
        "    <tr>\n",
        "      <td>0</td>\n",
-       "      <td>0.288129</td>\n",
-       "      <td>0.223384</td>\n",
-       "      <td>0.069689</td>\n",
+       "      <td>0.293625</td>\n",
+       "      <td>0.205537</td>\n",
+       "      <td>0.067659</td>\n",
        "      <td>00:27</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <td>1</td>\n",
-       "      <td>0.204079</td>\n",
-       "      <td>0.197346</td>\n",
-       "      <td>0.067659</td>\n",
+       "      <td>0.195267</td>\n",
+       "      <td>0.185939</td>\n",
+       "      <td>0.055480</td>\n",
        "      <td>00:27</td>\n",
        "    </tr>\n",
        "    <tr>\n",
        "      <td>2</td>\n",
-       "      <td>0.131804</td>\n",
-       "      <td>0.183479</td>\n",
-       "      <td>0.058863</td>\n",
+       "      <td>0.123829</td>\n",
+       "      <td>0.172681</td>\n",
+       "      <td>0.055480</td>\n",
        "      <td>00:27</td>\n",
        "    </tr>\n",
        "  </tbody>\n",