Saving local graphs and readme

.idea/.gitignore

@@ -0,0 +1,10 @@
+# Default ignored files
+/shelf/
+/workspace.xml
+# Ignored default folder with query files
+/queries/
+# Datasource local storage ignored files
+/dataSources/
+/dataSources.local.xml
+# Editor-based HTTP Client requests
+/httpRequests/

.idea/document-classification.iml

@@ -0,0 +1,8 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<module type="PYTHON_MODULE" version="4">
+  <component name="NewModuleRootManager">
+    <content url="file://$MODULE_DIR$" />
+    <orderEntry type="inheritedJdk" />
+    <orderEntry type="sourceFolder" forTests="false" />
+  </component>
+</module>

.idea/inspectionProfiles/profiles_settings.xml

@@ -0,0 +1,6 @@
+<component name="InspectionProjectProfileManager">
+  <settings>
+    <option name="USE_PROJECT_PROFILE" value="false" />
+    <version value="1.0" />
+  </settings>
+</component>

.idea/modules.xml

@@ -0,0 +1,8 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+  <component name="ProjectModuleManager">
+    <modules>
+      <module fileurl="file://$PROJECT_DIR$/.idea/document-classification.iml" filepath="$PROJECT_DIR$/.idea/document-classification.iml" />
+    </modules>
+  </component>
+</project>

README.md

@@ -38,8 +38,6 @@ This model is a **ResNet-50** Convolutional Neural Network (CNN) finetuned to cl
 
 ## Model Details
 
-![Model Architecture](aechitecture.png)
-
 ### Model Description
 
 This model utilizes the standard ResNet-50 architecture designed for image classification. Instead of "reading" the text like an OCR system, it analyzes the visual layout, structure, and low-level texture features of a whole document page to determine its category (e.g., recognizing the block layout of a resume versus the dense, two-column text of a scientific report).
@@ -50,20 +48,7 @@ It was trained using **Transfer Learning**, starting with weights pre-trained on
 - **Model type:** Computer Vision (Image Classification / CNN)
 - **Language(s) (NLP):** English (Implicitly, via the text present in the RVL-CDIP dataset images)
 - **License:** MIT
-
-## Why ResNet50
-
-| Model      | Approximate Parameters | Year Released | Layers |
-|------------|------------------------|---------------|--------|
-| VGG16      | 138.4 Million          | 2014          | 16     |
-| AlexNet    | 61.1 Million           | 2012          | 8      |
-| ResNet-50  | 25.6 Million           | 2015          | 50     |
-
-| Model      | FLOPs (Billions) | Efficiency Score      |
-|------------|------------------|-----------------------|
-| AlexNet    | 0.7 GFLOPs       | Low Cost / Low Acc    |
-| ResNet-50  | 3.8 GFLOPs       | High Efficiency       |
-| VGG-16     | 15.5 GFLOPs      | Terribly Inefficient  |
+- **Finetuned from model:** ResNet-50 (ImageNet weights)
 
 ### Model Sources
 
@@ -198,15 +183,6 @@ The model was evaluated on the standard, unseen **RVL-CDIP Test Split** containi
 | **Overall Accuracy** | **88.46%** | Solid baseline performance. |
 | **Top-3 Accuracy** | **95.62%** | Excellent reliability for triage tasks. |
 
-![Loss and Accuracy Curves](results/loss_and_acc_curve.png)
-
-#### Confusion Matrix
-![Confusion Matrix](results/cm.png)
-
-#### Detailed Classificatio report
-![Detailed Classification report](results/detailed_classification_report.png)
-
-
 #### Detailed Performance Analysis (The "Traffic Light" Report)
 
 An analysis of per-class F1-scores reveals distinct tiers of performance: