Dataset

This dataset was created by Ashifur Rahman

Contents

DFDC-10: Deepfake Detection Training Dataset

Overview

This dataset contains the first 10 chunks (00.zip - 09.zip) from the Deepfake Detection Challenge (DFDC) dataset, totaling approximately 100GB of high-quality video data. Specifically curated for training specialist deepfake detection models with balanced real and manipulated video samples.

Dataset Structure

• Total Size: ~100GB
• Chunks: 10 files (00.zip through 09.zip)
• Chunk Size: ~10GB each
• Video Format: MP4, AVI, MOV
• Resolution: Various (480p to 1080p)
• Duration: 2-10 seconds per video
• Labels: Binary (real/fake) with metadata

Content Distribution

• Real Videos: ~50% authentic, unmanipulated content
• Fake Videos: ~50% deepfake manipulations using various techniques
• Manipulation Types: Face swapping, facial reenactment, speech-driven animation
• Quality Levels: High, medium, and low compression artifacts
• Lighting Conditions: Normal, low-light, varied environments

Use Cases

• Training deepfake detection models
• Transfer learning for video analysis
• Specialist model development (compression, temporal, audio-visual analysis)
• Research in synthetic media detection
• Benchmark testing for detection algorithms

Technical Specifications

• Video Codecs: H.264, H.265
• Audio: AAC, MP3 (where available)
• Frame Rates: 24-30 FPS
• Aspect Ratios: 16:9, 4:3, various mobile formats
• Color Space: RGB, YUV420p

Training Recommendations

• Batch Size: 8-16 depending on GPU memory
• Input Resolution: 224x224 for efficient training
• Augmentation: Rotation, color jitter, horizontal flip
• Architecture: EfficientNet-B4 backbone recommended
• Transfer Learning: Use ImageNet pretrained weights

Data Processing Notes

• Videos are stored in compressed ZIP archives
• Extract sequentially for incremental training
• Balanced sampling recommended to avoid bias
• Frame extraction at 1-2 FPS sufficient for most models

Citation

If you use this dataset in your research, please cite: - Original DFDC Challenge dataset - This curated subset for specialist model training

License

Please refer to the original DFDC dataset license terms.

Compatibility

• Frameworks: PyTorch, TensorFlow, OpenCV
• Platforms: Kaggle, Google Colab, local training
• Hardware: GPU recommended (T4, P100, V100)
• Memory: 16GB+ RAM recommended for efficient processing

Quality Assurance

• All videos manually verified for quality
• Balanced real/fake distribution maintained
• Corrupted files removed
• Consistent labeling across all chunks

Perfect for training robust deepfake detection models with state-of-the-art performance!

dfb-data/dfdc dataset hosted on Hugging Face and contributed by the HF Datasets community

The DFDC (Deepfake Detection Challenge) is a dataset for deepface detection consisting of more than 100,000 videos. The DFDC dataset consists of two versions: Preview dataset. with 5k videos. Featuring two facial modification algorithms. Full dataset, with 124k videos. Featuring eight facial modification algorithms

This dataset contains a very small subset of dfdc dataset, with 2203 videos and related metadata

Metadata columns : Index(['filename', 'video.@index', 'video.@codec_time_base', 'video.@width', 'video.@height', 'video.@sample_aspect_ratio', 'video.@display_aspect_ratio', 'video.@level', 'video.@r_frame_rate', 'video.@time_base', 'video.@duration_ts', 'video.@duration', 'video.@bit_rate', 'video.@nb_frames', 'audio.@index', 'audio.@codec_name', 'audio.@codec_long_name', 'audio.@codec_time_base', 'audio.@duration_ts', 'audio.@duration', 'audio.@bit_rate', 'audio.@max_bit_rate', 'audio.@nb_frames', 'label', 'split', 'original', 'folder', 'wav.hash.cnt', 'original.cnt', 'md5', 'md5.orig', 'wav.hash', 'wav.hash.orig', 'pxl.hash', 'pxl.hash.orig'], dtype='object')

laodeAlif/dfdc dataset hosted on Hugging Face and contributed by the HF Datasets community

dfb-data/dfdc-faces dataset hosted on Hugging Face and contributed by the HF Datasets community

Dataset

This dataset was created by Fakecatcher AI

Released under CC0: Public Domain

Contents

The DFDC dataset contains 100,000 images of faces manipulated using Deepfakes.

Dataset

This dataset was created by Khafie0212

Released under CC0: Public Domain

Contents

Overview

In Deepfake Detection Challenge, many public notebooks use the faces at equidistant frames of the videos. Here are all the faces of the test sample.

Feel free to use this dataset to test your models in Deepfake Detection Challenge. If you find this dataset useful, please consider upvoting it.

Creation Method

The videos were first used to extract keyframes(limited to Max 20 frames from each video) from them, and then the faces were detected and cropped from the extracted frames with the highest confidence (>= 95%) value.

The notebook used to create the image dataset from video samples can be found here

DFDC Faces MTCNN

Dossier généré avec 5 frames par vidéo.

150x150 face images from some frames from every video in part 18 of Deepfake Detection train data set. Specifically 10 frames evenly taken from all parts of every video.

View Camping world dfdc import data USA including customs records, shipments, HS codes, suppliers, buyer details & company profile at Seair Exim.

Dataset

This dataset was created by Mohasina Jannat Moon

Contents

Explore the historical Whois records related to dfdc.info (Domain). Get insights into ownership history and changes over time.

Comparison between different combinations of Mixformer. The results in the table are test with the DFDC dataset (in %).

The paper proposes a unified face forgery detection framework to solve the multi-dataset conflict problem, and the authors use five public autopilot datasets, including FaceForensics++ [14] Celeb-DF(V2) [15], WildDeepfake [16], DFDC [17] and the fake face dataset generated by diffusion DFF [18] to study the problem of data conflict in each domain or merged domains.

Dataset

This dataset was created by Enoch Bidi

Released under CC0: Public Domain

Contents

Deepfake detection remains a pressing challenge, particularly in real-world settings where smartphone-captured media often introduces Moiré artifacts that can distort detection outcomes. This study systematically evaluates state-of-the-art (SOTA) deepfake detectors on Moiré-affected videos—an issue that has received little attention. We collected a dataset of 12,832 videos, spanning 35.64 hours, from CelebDF, DFD, DFDC, UADFV, and FF++ datasets, capturing footage under diverse real-world conditions, including varying screens, smartphones, lighting setups, and camera angles.

We provide the dataset request through the Google Form below:

https://forms.gle/oifqaoujH6q73JnR6">https://forms.gle/oifqaoujH6q73JnR6