GAMBIT

Dataset image markup program for training any neural networks on any objects 

With this program, you can create neural networks for yourself and others

Select the desired object, and you will receive an automatic reaction to it! 

07.05.2025   Current version - 1.19  Download DATA SET image partitioning program for neural networks under Windows

The entire process is designed for "dummies"
artificial intelligence is now accessible to people without education

Yes, there are many large companies that already sell neural networks that understand any objects. But neither OpenAI, nor Samsung, nor Microsoft will specifically train a neural network for you so that it lets your specific dog into your house, so that it records the readings of your specific meter, so that it monitors the arrow on your specific device, so that it monitors the clogging of storm drains specifically at your house (otherwise the basement will flood), so that it tracks the quality of products at your specific production, so that it searches for defects on your specific machine, so that it monitors compliance with the rules of work of your specific employees… This list is endless – people have so many unique tasks in the field of AI.

Neither OpenAI, nor Samsung, nor Microsoft will retrain the neural network for you if you specifically have false positives or no positive recognitions.

But not everyone has the money for top-notch programmers. Not everyone even understands how it works. Speclab provides a simple, understandable tool for creating AI, accessible to everyone. And you can try it right now.

It’s simple: take an ordinary video camera, or a mobile phone, or any video surveillance system, shoot a video clip of the object or process you would like to control. Then place this video clip into the “Gambit” program and highlight the objects you need there. After going through a certain number of video frames, you get what is called a DATASET, which is the basis for training any neural network.

Upload it to our AI cloud, where it transforms into a ready-to-use neural network. Moreover, immediately with service functions. For example, the system can send you an email if your favorite flower starts to wilt. Or close the Smart Home relay if the camera detects a puddle of water on the floor. Or… a million variations, it’s a real neural network factory.

Although… this program was originally created for much more than such elementary applications. With its help, you can create datasets for detecting serious processes and a wide variety of searched objects. For example, the same abandoned objects have hundreds of thousands of different types, from bags and packages to boxes and suitcases. And here, we need manpower, lots and lots of labeled pictures. The software from Speclab allows you to load “remote workers” with work, including for the needs of Speclab. Those interested are welcome!

In this case, of course, not everything is so simple. You need attentiveness and perseverance to carefully highlight the given objects in large quantities. These qualities do not relate to education or even to a field of knowledge; people who know how to knit, for example, or perform other constantly repeating work would be a good fit. Therefore, we do not take everyone, but only those who do not make mistakes. But you can try, we give three attempts. In connection with the liberalization of laws, we also accept children from 14 years old. But in all cases, the candidate must have basic computer literacy, such as how to send an email, download, or upload a file to the cloud. We won’t teach you how to tie your shoelaces.

“Gambit” is not so much a commercial product as it is a supplementary one. We ourselves are constantly engaged in training neural networks, and we haven’t found anything convenient and accessible on the Internet, despite the seemingly mass reproduction of neural networks. In the vast majority of cases, it’s a large garbage dump; we’ll talk more about their quality later. But even if you compare our program with projects not intended for everyone (you can’t download them for free and you can’t always find them on the Internet, because large companies hide their strategic potential), like AI-Markup, VisionAnnotator, ObjectMapper, SegmentPro, SegmentMaster, all of this seemed inconvenient and flawed to us. And most importantly, there’s no feedback there; the quality of the markup isn’t controlled in any way; people work blindly. Not to mention self-learning and other useful features.

Speclab created Gambit primarily for itself and uses it for a large number of its products based on training neural networks for video surveillance systems, from video blazers to GOAlcity. This markup system is constantly being tested on our employees, and a new version is released almost every week – even smarter, even more convenient.

We’d be happy to hear feedback and suggestions for new ideas in this area! Because we ourselves need it. The basic functionality is free – use it for the sake of AI! Only the high-level piloting functions that we have already begun to insert for training neural networks will be paid. One of them already exists – and it eliminates false positives. We haven’t seen anything like this in the software of OpenAI or other AI giants. If their neural network screws up something, there’s no way to fix it. And it is false positives that kill any good idea.

It is the quality of the dataset that determines how effectively the neural network will work. Speclab’s vast experience is invested in the methodology of creating metadata. It’s described in both the help documentation and in training videos. We also conduct training seminars for everyone interested. Everything is simple and accessible, an end to the wise incomprehensibilities in AI!