How to evaluate Glykon on your own dataset

test_dataset
├── fake
│   └── video1.mp4
├── real
    └── video1.mp4

from sklearn.metrics import classification_report
from imutils import paths
import numpy as np
import argparse
import cv2
from sklearn.metrics import classification_report, confusion_matrix, roc_curve
import plotly.graph_objects as go
import plotly
import requests

#function to call Glykon API
def call_glykon(file):
    base = "http://localhost:8000"
    url = "%s/run" % base
    data = {
            "video":open(file, "rb")
    }
    r = requests.post(url, files=data).json()
    return r

#Load data in test dataset
data = paths.list_files("ground_truths")

#Loop over data processing it with Glykon and saving results
y_true = []
scores = []
y_pred = []
for file in data:
    l_true = file.split("/")[-2]
    print(l_true)
    score = call_glykon(file)
    print(score)
    if score["prediction"] == None:
        print("Video Rejected!")
        print(score["guidance"])
        continue
    scores.append(score["frame_analysis_real_score"])
    y_pred.append(score["prediction"])
    y_true.append(l_true)

#encode labels
y_true = [0 if x == "fake" else 1 for x in y_true]
y_pred = [0 if x == "fake" else 1 for x in y_pred]


#Generate Confusion Matrix
cm = confusion_matrix(
    y_true, 
    y_pred)

#Measure Accuracy
l = len(y_true)
acc = sum([y_pred[i]==y_true[i] for i in range(l)])/l
target_names = ['real', 'fake']

#Generate Classification Report
cr = classification_report(y_true, y_pred, target_names=target_names)

fpr, tpr, thresholds = roc_curve(
    y_true, 
    scores, pos_label=1)
print("fpr: %s" % fpr)
print("tpr: %s" % tpr)

thresholds = [str(t) for t in thresholds]
fig = go.Figure(data=go.Scatter(x=fpr, y=tpr, hovertext=thresholds))
plot_name = "roc.html"
fig.write_html(plot_name, auto_open=True)

print("Accuracy: %s\nConfusion Matrix:\n%s\nClassification Report:\n%s" % (acc, cm, cr))
print("ROC Curve plot saved to %s" % plot_name)

Accuracy: 0.8095238095238095
Confusion Matrix:
[[8 0]
 [4 9]]
Classification Report:
               precision    recall  f1-score   support

        real       0.67      1.00      0.80         8
        fake       1.00      0.69      0.82        13

    accuracy                           0.81        21
   macro avg       0.83      0.85      0.81        21
weighted avg       0.87      0.81      0.81        21

ROC Curve plot saved to roc.html