combine detectors into a single API

.gitignore

@@ -1,4 +1,5 @@
 .env
+.venv
 dist
 .eggs
 *.egg-info

detectors/Dockerfile

@@ -0,0 +1,7 @@
+FROM nvidia/cuda:11.0-base
+RUN apt-get update && apt-get install -y python3-pip  wget
+WORKDIR /
+COPY . /
+RUN python3 -m pip install -U pip && python3 -m pip install -r requirements.txt
+EXPOSE 5000
+CMD ["python3","app.py"]

detectors/app.py

@@ -0,0 +1,64 @@
+import os
+import pandas as pd
+import pickle
+import flask
+from flask import Flask, request, jsonify, make_response
+from boken.ensemble import Ensemble as boken_Ensemble
+from eighteen.ensemble import Ensemble as eighteen_Ensemble
+from medics.ensemble import Ensemble as medics_Ensemble
+from ntech.ensemble import Ensemble as ntech_Ensemble
+from selimsef.ensemble import Ensemble as selimsef_Ensemble
+from wm.ensemble import Ensemble as wm_Ensemble
+import boto3
+
+app = Flask(__name__)
+
+models = {
+    "boken": boken_Ensemble,
+    "eighteen": eighteen_Ensemble,
+    "medics": medics_Ensemble,
+    "ntech": ntech_Ensemble,
+    "selimsef": selimsef_Ensemble,
+    "wm": wm_Ensemble,
+}
+
+
+@app.route('/healthcheck')
+def starting_url():
+    status_code = flask.Response(status=201)
+    return status_code
+
+@app.route('/predict', methods=['POST'])
+def predict():
+    model_names = request.get_json(force=True)['model_names']
+    video_list = request.get_json(force=True)['video_list']
+    predictions = []
+    for filename in video_list:
+        for modelname in model_names:
+            score = 0.5
+            video = ''
+            model = models[modelname]()
+            try:
+                validate_filename(filename)
+                video = sanitize_filename(filename, platform="auto")
+                video_path = os.path.join('/uploads/', video)
+                if os.path.exists(video_path):
+                    score = model.inference(video_path)
+                    pred={'filename': video}
+                    pred[modelname]=score
+                else:
+                    return make_response(f"File {video} not found.", 400)
+            except ValidationError as e:
+                print(f'{e}')
+                return make_response(f"{e}", 400)
+            except Exception as err:
+                print(f'{err}')
+                return make_response(f"{err}", 500)
+
+    result = pd.DataFrame(predictions)
+    return result.to_json()
+
+
+
+if __name__ == '__main__':
+    app.run(host='0.0.0.0')

detectors/boken/ensemble.py

@@ -1,11 +1,11 @@
 import os
 import torch
 
-from eval_kit.detector import DeeperForensicsDetector
-from eval_kit.client_dev import get_local_frames_iter
-from models import model_selection, get_efficientnet, DetectionPipeline
+from .eval_kit.detector import DeeperForensicsDetector
+from .eval_kit.client_dev import get_local_frames_iter
+from .models import model_selection, get_efficientnet, DetectionPipeline
 from facenet_pytorch import MTCNN, extract_face
-from utils import *
+from .utils import *
 
 
 class Ensemble(DeeperForensicsDetector):

detectors/boken/eval_kit/client_dev.py

@@ -14,7 +14,7 @@
 
 from io import BytesIO
 
-from eval_kit.extract_frames import extract_frames
+from .extract_frames import extract_frames
 
 # EVALUATION SYSTEM SETTINGS
 # DON'T CHANGE ANY CODE HERE, OR YOU MAY ENCOUNTER SOME UNEXPECTED PROBLEMS!

detectors/eighteen/ensemble.py

@@ -8,12 +8,12 @@
 import torch.backends.cudnn as cudnn
 import torchvision
 
-from face_detect_lib.models.retinaface import RetinaFace
-from face_detect_lib.layers.functions.prior_box import PriorBox
-from face_detect_lib.utils.box_utils import decode_batch, decode_landm_batch, decode, decode_landm
+from .face_detect_lib.models.retinaface import RetinaFace
+from .face_detect_lib.layers.functions.prior_box import PriorBox
+from .face_detect_lib.utils.box_utils import decode_batch, decode_landm_batch, decode, decode_landm
 
-from utils import *
-from models import *
+from .utils import *
+from .models import *
 
 __all__ = ['Ensemble', 'pipeline_cfg']
 

detectors/eighteen/face_detect_lib/layers/modules/multibox_loss.py

@@ -2,8 +2,8 @@
 import torch.nn as nn
 import torch.nn.functional as F
 from torch.autograd import Variable
-from face_detect_lib.utils.box_utils import match, log_sum_exp
-from face_detect_lib.models.config import cfg_mnet
+from eighteen.face_detect_lib.utils.box_utils import match, log_sum_exp
+from eighteen.face_detect_lib.models.config import cfg_mnet
 GPU = cfg_mnet['gpu_train']
 
 class MultiBoxLoss(nn.Module):

detectors/eighteen/face_detect_lib/models/retinaface.py

@@ -5,9 +5,9 @@
 import torch.nn.functional as F
 from collections import OrderedDict
 
-from face_detect_lib.models.net import MobileNetV1 as MobileNetV1
-from face_detect_lib.models.net import FPN as FPN
-from face_detect_lib.models.net import SSH as SSH
+from .net import MobileNetV1 as MobileNetV1
+from .net import FPN as FPN
+from .net import SSH as SSH
 
 
 class ClassHead(nn.Module):

detectors/medics/ensemble.py

@@ -6,10 +6,10 @@
 from torchvision.models.video import mc3_18, r2plus1d_18
 from facenet_pytorch import MTCNN
 
-from cnn3d import *
-from cnn2d import *
-from constants import *
-from face_utils import *
+from .cnn3d import *
+from .cnn2d import *
+from .constants import *
+from .face_utils import *
 
 __all__ = ['Ensemble']
 

detectors/medics/face_utils.py

@@ -3,7 +3,7 @@
 import cv2
 import math
 import skimage.measure
-from constants import *
+from .constants import *
 
 
 def load_video(filename, every_n_frames=None, specific_frames=None, to_rgb=True, rescale=None, inc_pil=False, max_frames=None):

detectors/ntech/ensemble.py

@@ -1,8 +1,8 @@
 import numpy as np
 import cv2
 
-from face_utils import get_tracks, extract_sequence, extract_face
-from models import *
+from .face_utils import get_tracks, extract_sequence, extract_face
+from .models import *
 
 VIDEO_FACE_MODEL_TRACK_STEP = 2
 VIDEO_SEQUENCE_MODEL_SEQUENCE_LENGTH = 7

detectors/ntech/face_utils.py

@@ -1,7 +1,7 @@
 import numpy as np
 import cv2
 
-from tracker.iou_tracker import track_iou
+from .tracker.iou_tracker import track_iou
 
 DETECTOR_STEP = 3
 

detectors/requirements.txt

@@ -0,0 +1,89 @@
+pathvalidate==2.5.0
+numpy==1.18.1
+opencv_python==4.2.0.34
+torch==1.4.0
+torchvision==0.5.0
+albumentations==0.5.2
+efficientnet_pytorch==0.7.0
+facenet_pytorch==2.5.1
+Pillow==8.1.1
+pretrainedmodels==0.7.4
+Flask==1.1.2
+easydict==1.8
+PyYAML==5.3.1
+tensorboardX==2.0
+pandas==1.0.4
+pydocs==0.2
+boto3==1.17.5
+scipy==1.4.1
+scikit_image==0.18.1
+Cython==0.29.21
+decord==0.4.2
+einops==0.3.0
+lmdb==1.0.0
+scikit_learn==0.24.1
+timm==0.3.4
+appdirs==1.4.3
+attrs==19.3.0
+autopep8==1.5.2
+backcall==0.1.0
+bleach==3.1.4
+certifi==2020.4.5.1
+chardet==3.0.4
+cycler==0.10.0
+decorator==4.4.2
+defusedxml==0.6.0
+distlib==0.3.0
+docopt==0.6.2
+entrypoints==0.3
+filelock==3.0.12
+flake8==3.7.9
+idna==2.9
+imageio==2.9.0
+importlib-metadata==1.5.0
+jedi==0.16.0
+Jinja2==2.11.1
+jsonschema==3.2.0
+mccabe==0.6.1
+mistune==0.8.4
+nbconvert==5.6.1
+nbformat==5.0.4
+networkx==2.5
+notebook==6.0.3
+olefile==0.46
+pandocfilters==1.4.2
+parso==0.6.2
+pexpect==4.8.0
+pickleshare==0.7.5
+pipreqs==0.4.10
+prometheus-client==0.7.1
+prompt-toolkit==3.0.4
+ptyprocess==0.6.0
+pycodestyle==2.5.0
+pyflakes==2.1.1
+Pygments==2.6.1
+pyparsing==2.4.6
+pyrsistent==0.16.0
+python-dateutil==2.8.1
+pytz==2019.3
+PyWavelets==1.1.1
+pyzmq==18.1.1
+requests==2.23.0
+scikit-image==0.18.1
+Send2Trash==1.5.0
+sip
+six==1.14.0
+terminado==0.8.3
+testpath==0.4.4
+tifffile==2021.1.14
+
+tornado==6.0.4
+tqdm==4.44.1
+traitlets==4.3.3
+urllib3==1.25.9
+virtualenv==20.0.18
+wcwidth==0.1.9
+webencodings==0.5.1
+widgetsnbextension==3.5.1
+yarg==0.1.9
+zipp==2.2.0

detectors/selimsef/ensemble.py

@@ -1,8 +1,8 @@
 import re
 
 import torch
-from kernel_utils import VideoReader, FaceExtractor, confident_strategy, predict_on_video
-from models.classifiers import DeepFakeClassifier
+from .kernel_utils import VideoReader, FaceExtractor, confident_strategy, predict_on_video
+from .models.classifiers import DeepFakeClassifier
 
 model_chpt = ['final_111_DeepFakeClassifier_tf_efficientnet_b7_ns_0_36',
               'final_555_DeepFakeClassifier_tf_efficientnet_b7_ns_0_19',

detectors/wm/ensemble.py

@@ -12,9 +12,9 @@
 from torchvision import transforms as T
 
 sys.path.append('./external/Pytorch_Retinaface')
-from model_def import WSDAN, xception
-from face_utils import norm_crop, FaceDetector
-from external.Pytorch_Retinaface.data import cfg_re50
+from .model_def import WSDAN, xception
+from .face_utils import norm_crop, FaceDetector
+from .external.Pytorch_Retinaface.data import cfg_re50
 
 class video_reader:
     def __init__(self, face_detector, transform=None, frame_skip=9, face_limit=25, batch_size=25):

detectors/wm/face_utils.py

@@ -4,10 +4,10 @@
 import torch
 from skimage.transform import SimilarityTransform
 
-from data import cfg_re50
-from layers.functions.prior_box import PriorBox
-from utils.box_utils import decode, decode_landm
-from models.retinaface import RetinaFace
+from .external.Pytorch_Retinaface.data import cfg_re50
+from .external.Pytorch_Retinaface.layers.functions.prior_box import PriorBox
+from .external.Pytorch_Retinaface.utils.box_utils import decode, decode_landm
+from .external.Pytorch_Retinaface.models.retinaface import RetinaFace
 
 
 def norm_crop(img, landmark, image_size=112):

detectors/wm/model_def/wsdan.py

@@ -13,7 +13,7 @@
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
-from .efficientnet_pytorch import EfficientNet as ef
+from efficientnet_pytorch import EfficientNet as ef
 from .xception_wsdan import xception
 
 __all__ = ['WSDAN']