Spot List of News Articles

We have released three technical documents for the AI visual inspection platform "DeepAge Sensor" for free. The document titled "How to Launch Visual Inspection AI in a Process with No Defective Samples" (12 pages) summarizes five steps to establish anomaly detection using only good product images, along with pitfalls identified through real data validation. Additionally, we have prepared a product catalog (14 pages) and a "Guide to Integrating AI into Inspection Equipment" (13 pages) for equipment manufacturers. You can download them by providing your name, company name, and email address.

We have started offering a desktop application called "Sensor AD" that learns the "normal standard" only from good product images and detects images that deviate from this standard as anomalies. There is no need to prepare a large number of defective product samples, allowing for the implementation of anomaly detection operations from the initial stages even in environments where collecting defective products is difficult. The inference runs on the CPU, and a GPU is not required. It features visualization through contribution heatmaps and automatic clustering of false positives.

Revamped the DeepAge Sensor's Muon optimizer with Frobenius normalization and BF16 numerical contracts. Validation on 20 real data sets achieved a median maximum accuracy of 100%, with 90% accuracy reached approximately 2.6 times faster than SGD (600 batches vs 1,700 batches). Achieved 97-100% accuracy on high-difficulty data sets that could not be learned with SGD. Operation on the latest NVIDIA GPUs (Blackwell/sm_120) has also been verified.

Released Early Stopping feature for DeepAge Sensor. Automatically halts training when the validation accuracy reaches the target value, reducing training time by up to half while maintaining accuracy. Efficient use of GPU resources allows for training more models in a shorter time.

Added fine-tuning (additional learning) functionality to the DeepAge Sensor. It is now possible to retrain using only additional images based on the existing model, allowing for accuracy improvements with less data and shorter time compared to starting from scratch. The improvement cycle when new defect patterns emerge is significantly shortened.