-
-
- ← Back to Results
-
- 🔍 Clip Analysis
-{{ clip.clip_name }}
-
-
-
-
- 🎬 Video Playback
-
-
-
-
- 📹 Video Information:
- • Original file: {{ clip.filename }}
- • This is the exact clip analyzed by the model
- • Model processes 16 frames at 224x224 resolution -
- - • Original file: {{ clip.filename }}
- • This is the exact clip analyzed by the model
- • Model processes 16 frames at 224x224 resolution -
-
🎯 Model Prediction
- -
-
-
- {{ "%.3f"|format(clip.normalized_score) }}
- {{ clip.confidence.category }}
- {{ clip.confidence.label }}
-
-
- {{ (clip.normalized_score * 100)|round }}%
+
+
+
+
+ ← Back to Results
+
+ Clip Analysis
+{{ clip.clip_name }}
+
+
-
-
+
-
+
+
Video Playback
+
+
+
+
+ Video Information:
+ • Original file: {{ clip.filename }}
+ • This is the exact clip analyzed by the model. +
+ + • Original file: {{ clip.filename }}
+ • This is the exact clip analyzed by the model. +
- 🧮 Probability Estimate:
- Using sigmoid transformation: {{ "%.1f"|format(clip.probability * 100) }}%
-
- - Using sigmoid transformation: {{ "%.1f"|format(clip.probability * 100) }}%
-
P = 1 / (1 + e^(-logit))
- * Experimental feature based on estimated logit reconstruction
+
+
+
+
+ Model Prediction
+
+
+ {{ "%.3f"|format(clip.normalized_score) }}
+ {{ clip.confidence.category }}
+ {{ clip.confidence.label }}
+
+
+
+
+ Probability Estimate: {{ "%.1f"|format(clip.probability * 100) }}%
+
+ P = 1 / (1 + e^(-logit))
+
-
- {% if clip.frame_paths and clip.frame_paths|length > 0 %}
- 🔧 Technical Analysis
- -
- Clip ID:
- #{{ clip.id + 1 }}
-
-
-
- Normalized Score:
- {{ "%.6f"|format(clip.normalized_score) }}
-
-
-
- Estimated Probability:
- {{ "%.4f"|format(clip.probability) }} ({{ "%.1f"|format(clip.probability * 100) }}%)
-
-
-
- Confidence Level:
- {{ clip.confidence.level }}/5
-
-
-
- Category:
- {{ clip.confidence.category }}
-
-
-
- File Path:
- {{ clip.video_path }}
-
-
-
🖼️ Analyzed Frames (Adaptive Center Crop)
-- These are the exact 16 frames that were analyzed by the model, extracted from frames 85-100 (killshot moment) with an adaptive center crop. - Crop size is scaled based on video resolution to maintain the same field of view as the training data (6.25% of video width). - The model internally resizes these to 224x224 during processing. -
-
- {% for i in range(clip.frame_paths|length) %}
-
- 
-
- {% endif %}
-
- Frame {{ i + 1 }}/16
- {{ "%.3f"|format((i / 15.0)) }}s
+
+
-
+
- {% endfor %}
Technical Analysis
+-
+
- Clip ID: #{{ clip.id + 1 }} +
- Normalized Score: {{ "%.6f"|format(clip.normalized_score) }} +
- Estimated Probability: {{ "%.4f"|format(clip.probability) }} ({{ "%.1f"|format(clip.probability * 100) }}%) +
- Confidence Level: {{ clip.confidence.level }}/5 +
- Category: {{ clip.confidence.category }} +
- File Path: {{ clip.video_path }} +
-
- 💡 Tip: Click on any frame to view it in full size.
- These frames were extracted at 16 FPS and represent approximately 1 second of gameplay.
-
-
-
-
📊 Score Interpretation Guide
-
-
- 0.8 - 1.0
- Very High Confidence
- Likely Cheating -
- - Very High Confidence
- Likely Cheating -
- 0.6 - 0.8
- High Confidence
- Possible Cheating -
- - High Confidence
- Possible Cheating -
- 0.4 - 0.6
- Medium Confidence
- Uncertain -
- - Medium Confidence
- Uncertain -
- 0.2 - 0.4
- Low Confidence
- Likely Legitimate + {% if clip.frame_paths and clip.frame_paths|length > 0 %} +
- Low Confidence
- Likely Legitimate + {% if clip.frame_paths and clip.frame_paths|length > 0 %} +
+
-
+
- Analyzed Frames
+These are the 16 frames analyzed by the model, extracted from the killshot moment with an adaptive center crop.
+
+ {% for i in range(clip.frame_paths|length) %}
+
+
+ {% endfor %}
+
+
+
+ {{ "%.2f"|format((i / 15.0)) }}s
+
- 0.0 - 0.2
- Very Low Confidence
- Likely Legitimate +
+ {% endif %}
+
+ - Very Low Confidence
- Likely Legitimate +
+
-
+
Score Interpretation Guide
+
+
0.8-1.0
Very High
+ Very High
0.6-0.8
High
+ High
0.4-0.6
Medium
+ Medium
0.2-0.4
Low
+ Low
0.0-0.2
Very Low
+ Very Low
- ⚠️ Important Notes:
- • These are model predictions, not definitive proof
- • Always consider multiple factors when evaluating gameplay
- • Model trained on specific patterns and may not catch all cheating methods
- • Use results as guidance, not absolute truth -
-
- - • These are model predictions, not definitive proof
- • Always consider multiple factors when evaluating gameplay
- • Model trained on specific patterns and may not catch all cheating methods
- • Use results as guidance, not absolute truth -
-
🧠 Model Information
- -
- Architecture:
- Vision Transformer (ViT) with VideoMAE
+ Evaluation Progress
-
+
+
-
+ Important Note: Model predictions are not definitive proof. Always use these results as guidance, not absolute truth.
-
- Model Size:
- vit_giant_patch14_224 (~1B parameters)
+
-
+
-
+
Model Information
+-
+
- Architecture: Vision Transformer (ViT) +
- Model Size: vit_giant_patch14_224 +
- Analysis Time: {{ metadata.evaluation_timestamp[:19].replace('T', ' ') }} +
- Model Path: {{ metadata.model_path }} +
- Input Processing:
- 16 frames @ 224x224 resolution
-
+
- Training Data:
- CS2 gameplay clips (cheater vs legitimate)
-
+
+
+
\ No newline at end of file
diff --git a/templates/evaluation.html b/templates/evaluation.html
index 16a46202..0525bc25 100644
--- a/templates/evaluation.html
+++ b/templates/evaluation.html
@@ -4,146 +4,49 @@
-
+ 🔍 Evaluation Progress
-Analyzing clips for cheating detection
-
+
+
+
+ 🔍 Evaluation Progress
+Analyzing clips for cheating detection
+
+ ← Back to Main
- ← Back to Main
+
-
-
+
-
+
+ Evaluation Configuration
+
+
+ Model: {{ model_name }}
+Data Path: {{ data_path }}
+Process ID: {{ process_id }}
+Started: {{ start_time }}
+
-
+ Evaluation Configuration
-
- Model: {{ model_name }}
-
-
- Data Path: {{ data_path }}
-
-
- Process ID: {{ process_id }}
-
-
- Started: {{ start_time }}
-
-
+
-
+
+ 🔄 Evaluation in progress...
+
+
+
+ Starting evaluation...
+
-
- 🔄 Evaluation in progress...
-
Starting evaluation...
-
-
+
Connecting to evaluation stream...