Upload train_classifier.ipynb

train_classifier.ipynb

@@ -339,7 +339,7 @@
     "    min_samples = counts.min()\n",
     "    # Calculate 2.0 times the minimum sample size, rounded down to the nearest integer\n",
     "    # target_samples = int(2.0 * min_samples)\n",
-    "    target_samples = 5000\n",
+    "    target_samples = 7500\n",
     "    \n",
     "    indices_to_keep = np.hstack([\n",
     "        np.random.choice(\n",
@@ -521,7 +521,7 @@
     "# Loss and optimizer\n",
     "criterion = nn.CrossEntropyLoss()\n",
     "optimizer = optim.Adam(model.parameters(), lr=0.0001, weight_decay=1e-5) \n",
-    "lambda_l1 = 1e-3  # L1 regularization strength"
+    "lambda_l1 = 1e-5  # L1 regularization strength"
    ]
   },
   {
@@ -539,7 +539,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "epochs = 50\n",
+    "epochs = 10\n",
     "train_losses, test_losses = [], []\n",
     "\n",
     "for epoch in range(epochs):\n",
@@ -577,7 +577,7 @@
     "    \n",
     "    precision, recall, f1, _ = precision_recall_fscore_support(all_targets, all_preds, average='weighted', zero_division=0)\n",
     "    accuracy = accuracy_score(all_targets, all_preds)  # Compute accuracy\n",
-    "    if epoch % 10==0:\n",
+    "    if epoch % 2==0:\n",
     "        print(f'Epoch {epoch+1}: Train Loss: {train_losses[-1]:.4f}, Test Loss: {test_losses[-1]:.4f}, Precision: {precision:.4f}, Recall: {recall:.4f}, F1: {f1:.4f}, Accuracy: {accuracy:.4f}')"
    ]
   },
@@ -620,6 +620,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "print(np.unique(all_targets, return_counts=True))\n",
+    "print(np.unique(all_preds, return_counts=True))\n",
+    "\n",
     "conf_matrix = confusion_matrix(all_targets, all_preds)\n",
     "labels = [\"background\", \"tackle-live\", \"tackle-replay\",]\n",
     "          # \"tackle-live-incomplete\", \"tackle-replay-incomplete\"]\n",
@@ -627,7 +630,19 @@
     "# plt.title('Confusion Matrix')\n",
     "plt.xlabel('Predicted Label')\n",
     "plt.ylabel('True Label')\n",
-    "plt.show()"
+    "plt.show()\n",
+    "\n",
+    "def showClassWiseAcc(conf_matrix):\n",
+    "    # Calculate accuracy per class\n",
+    "    class_accuracies = conf_matrix.diagonal() / conf_matrix.sum(axis=1)\n",
+    "\n",
+    "    # Prepare accuracy data for writing to file\n",
+    "    accuracy_data = \"\\n\".join([f\"Accuracy for class {i}: {class_accuracies[i]:.4f}\" for i in range(len(class_accuracies))])\n",
+    "\n",
+    "    # Print accuracy per class and write to a file\n",
+    "    print(accuracy_data)  # Print to console\n",
+    "\n",
+    "showClassWiseAcc(conf_matrix)"
    ]
   },
   {