AI Benchmark 6.0.0

AI Benchmark Professional Evaluation

Is your smartphone equipped to handle cutting-edge Deep Neural Networks for tasks like Face Recognition, Image Classification, and Question Answering? Does it boast a specialized AI Chip and the speed required for these operations? Evaluate your device’s AI Performance with AI Benchmark, a tool designed for professional analysis.

For the latest phone rankings, visit: AI Benchmark Ranking

AI Benchmark assesses various AI and Computer Vision algorithms based on speed, accuracy, power consumption, and memory usage. It tests Image Classification, Face Recognition techniques, Neural Networks for Image and Video Enhancement, as well as models for text prediction and question answering. Moreover, it evaluates AI systems utilized in autonomous driving and smartphones for tasks like Depth Estimation and Semantic Image Segmentation. Visual representations of algorithm outputs offer insight into the current state-of-the-art across diverse AI domains.

• Classification: MobileNet-V2
• Classification: Inception-V3
• Face Recognition: MobileNet-V3
• Classification: EfficientNet-B4
• Parallel Model Execution: 8 x Inception-V3
• Object Tracking: YOLO-V4
• Optical Character Recognition: CRNN
• Semantic Segmentation: DeepLabV3+
• Photo Deblurring: IMDN
• Image Super-Resolution: ESRGAN
• Image Denoising: U-Net
• Depth Estimation: MV3-Depth
• Image Enhancement: DPED ResNet
• Bokeh Effect Rendering: PyNET+
• Learned Camera ISP: PUNET
• FullHD Video Super-Resolution: XLSR
• 4K Video Super-Resolution: VideoSR
• Text Completion: LSTM
• Question Answering: MobileBERT
• Memory Limits: ResNet

In addition to the standard tests, users can assess their own TensorFlow Lite deep learning models using the PRO Mode feature. Detailed test descriptions can be accessed here: AI Benchmark Tests Description

Note: Hardware acceleration is available on mobile platforms with dedicated NPUs and AI accelerators including Qualcomm Snapdragon, HiSilicon Kirin, Samsung Exynos, MediaTek Helio / Dimensity, and UNISOC Tiger chipsets. With AI Benchmark v4 onwards, older devices can also enable GPU-based AI acceleration by toggling the setting in "Accelerate" -> "Enable GPU Acceleration" (requires OpenGL ES-3.0+).