On the 6th, Neurocle announced the release of Neuro-X. Neurocle has expanded its market from just vision software for industry specialists to vision software for deep learning experts, targeting all users who want to enhance their deep learning project experience.
Neuro-X offers various functions such as △modeling system that directly creates models optimized for various tasks and △data management system that reduces the resources of managing data.
The modeling system of Neuro-X provides model optimization without constraints. The possibilities for optimization are infinite without limits in setting different combinations of training hyperparameters to develop models. Compared to other existing software focused solely on data-based optimization (i.e., augmentation), Neurocle also provides model-based optimization. For example, by selecting different learning rates and learning rate decay options for each task, users can adjust the speed or objective of model training. Experts can fully exploit their deep learning expertise, experimenting with combinations of various conditions.
In addition, Neurocle's data management system is perfect for systematic data management of deep learning projects. Users can categorize data by various characteristics and take advantage of a structure suitable for history management. In fact, more than 80% of the average six months that deep learning researchers take to optimize and distribute a model is spent on tasks related to data. Neurocle software can reduce this data management resource by about 50 percent. Through this structure, users can also easily check the information about the project design, which can be a solution to the handover problem. Even when staffing changes occur, differences in research knowledge can be leveled off and the project can proceed smoothly.
“As deep learning vision technologies have been applied to various industries for tasks such as defect inspection in manufacturing, medical image interpretation, and logistics packaging inspection, there has been an increase in the number of users who have a rich knowledge in deep learning. Thus, we deemed it appropriate to release Neuro-X to suit the needs of these users who want to utilize their expert knowledge in a deep learning project," Neurocle official said, "The goal is to gradually reflect what experts have previously attempted with deep learning libraries such as TensorFlow in our products, and to provide a generalized environment where researchers and students majoring in related fields can conduct experiments with a high degree of freedom."
Comments