Multimodel_setup / scripts /test_univfd_adapter.py
Thiru1122's picture
Add standardized detector adapter and evaluation documentation
6f9e017
Raw History Blame Contribute Delete
6.96 kB
"""
Diagnostic Test Suite for UnivFDAdapter
Validates adapter integration, schema adherence, error handling, and input rejection.
NOTE: This script verifies operational pipeline behavior. It is NOT an accuracy evaluation
and does NOT verify image authenticity.
"""
import os
import sys
import json
from PIL import Image
# Add project root to sys.path
BASE_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
if BASE_DIR not in sys.path:
sys.path.insert(0, BASE_DIR)
from src.univfd_adapter import UnivFDAdapter
from src.detector_schema import PREDICTION_REAL, PREDICTION_FAKE, PREDICTION_ERROR
ASSET_DIR = r"C:\Users\user\.gemini\antigravity-ide\brain\3b54ff09-c09f-4fe6-9276-f4cca9fbc6d7"
def main():
print("=" * 75)
print("UNIVERSALFAKEDETECT ADAPTER DIAGNOSTIC TEST SUITE")
print("Verifies adapter schema, error handling, and inference integrity.")
print("DISCLAIMER: This test validates integration mechanics, not model accuracy.")
print("=" * 75 + "\n")
adapter = UnivFDAdapter(lazy_load=True)
# Prepare non-image text file for test case 5
invalid_txt_path = os.path.join(BASE_DIR, "scripts", "_temp_invalid.txt")
with open(invalid_txt_path, "w", encoding="utf-8") as f:
f.write("This is a plain text file, not an image format.")
test_cases = [
{
"id": 1,
"category": "VALID_REAL_IMAGE",
"name": "Known real photograph (smartphone capture)",
"input": os.path.join(ASSET_DIR, "scratch", "real_phone_photo.png"),
"ground_truth_label": "REAL",
"expected_outcome": "SUCCESSFUL_INFERENCE",
"nature": "Natural photographic capture from smartphone camera.",
},
{
"id": 2,
"category": "VALID_AI_IMAGE",
"name": "Known AI-generated image (Google Imagen 3)",
"input": os.path.join(ASSET_DIR, "ai_cyberpunk_city_1789660863192.jpg"),
"ground_truth_label": "FAKE",
"expected_outcome": "SUCCESSFUL_INFERENCE",
"nature": "Modern text-to-image diffusion generation (out-of-distribution for ProGAN-trained head).",
},
{
"id": 3,
"category": "VALID_MANIPULATED_IMAGE",
"name": "Known manipulated image (digital composite)",
"input": os.path.join(ASSET_DIR, "scratch", "manipulated_image.png"),
"ground_truth_label": "MANIPULATED",
"expected_outcome": "SUCCESSFUL_INFERENCE",
"nature": "Digitally spliced composite of photographic and synthetic elements.",
},
{
"id": 4,
"category": "CORRUPTED_FILE",
"name": "Corrupted image file (truncated header)",
"input": os.path.join(ASSET_DIR, "scratch", "corrupted_image.png"),
"ground_truth_label": "CORRUPTED",
"expected_outcome": "INPUT_REJECTION",
"nature": "Invalid 20-byte file that cannot be decoded by image decoders.",
},
{
"id": 5,
"category": "UNSUPPORTED_INPUT",
"name": "Unsupported file format (plain text document)",
"input": invalid_txt_path,
"ground_truth_label": "INVALID_TYPE",
"expected_outcome": "INPUT_REJECTION",
"nature": "Non-image text file submitted to image inference pipeline.",
},
]
all_results = []
for tc in test_cases:
print("-" * 75)
print(f"TEST {tc['id']}: {tc['name']}")
print(f" Input Source: {tc['input']}")
print(f" Category: {tc['category']}")
print(f" Ground-Truth Label: {tc['ground_truth_label']}")
print(f" Expected Outcome: {tc['expected_outcome']}")
print(f" Test Nature: {tc['nature']}")
result = adapter.predict(tc["input"])
res_dict = result.to_dict()
# Operational status assessment
is_error = result.prediction == PREDICTION_ERROR
if tc["expected_outcome"] == "INPUT_REJECTION":
status_classification = "INPUT_REJECTION_VERIFIED" if is_error else "FAILURE (Did not reject)"
else:
status_classification = "SUCCESSFUL_INFERENCE" if not is_error else "EXECUTION_FAILURE"
# Correctness vs Model Limitation assessment
if not is_error:
if result.prediction == tc["ground_truth_label"]:
domain_assessment = "ACCORDANT (Prediction matches ground truth label)"
else:
domain_assessment = (
"MODEL_LIMITATION_OBSERVED (Known ProGAN-trained classifier boundary: "
f"output {result.prediction} on {tc['ground_truth_label']} input)"
)
else:
domain_assessment = "N/A (Input correctly rejected before model evaluation)"
print(f" Prediction: {result.prediction}")
print(f" Raw Model Logit: {f'{result.raw_logit:+.5f}' if result.raw_logit is not None else 'N/A'}")
print(f" Fake Probability: {f'{result.fake_probability:.6f}' if result.fake_probability is not None else 'N/A'}")
print(f" Confidence: {f'{result.confidence*100:.2f}%' if result.confidence is not None else 'N/A'}")
print(f" Threshold: {result.threshold}")
print(f" Compute Device: {result.device}")
print(f" Inference Latency: {f'{result.inference_time_ms:.2f} ms' if result.inference_time_ms is not None else 'N/A'}")
print(f" Error Field: {result.error if result.error else 'None'}")
print(f" Operational Status: {status_classification}")
print(f" Domain Assessment: {domain_assessment}")
all_results.append({
"test_id": tc["id"],
"name": tc["name"],
"category": tc["category"],
"ground_truth_label": tc["ground_truth_label"],
"expected_outcome": tc["expected_outcome"],
"operational_status": status_classification,
"domain_assessment": domain_assessment,
"result": res_dict
})
print()
# Clean up temporary test file
if os.path.exists(invalid_txt_path):
try:
os.remove(invalid_txt_path)
except Exception:
pass
print("=" * 75)
print("TEST SUITE SUMMARY")
print("=" * 75)
for r in all_results:
res = r["result"]
print(f"Test {r['test_id']}: {r['name'][:30]:<30} | Pred: {res['prediction']:<6} | Logit: {str(res['raw_logit']):<9} | Op: {r['operational_status']}")
summary_file = os.path.join(BASE_DIR, "scripts", "adapter_test_results.json")
with open(summary_file, "w", encoding="utf-8") as f:
json.dump(all_results, f, indent=2)
print(f"\nSaved structured test results to: {summary_file}")
if __name__ == "__main__":
main()