Your external visual memory
At NeuralCam, we're really excited about the latest developments in AI technology. We believe that AI should enhance, not replace, human intelligence, and we're committed to developing software tools to bring a slice of this vision to life. Today, we're delighted to introduce our newest creation, NeuralBox. This AI-powered camera and gallery app is designed to give your visual memory a boost, helping you remember anything you see using photos.
The Imperfection of Human Memory
Every day, we're overwhelmed with information, and naturally, we forget most of it. This forgetting is often a blessing, a necessary mechanism for maintaining sanity in an information-saturated world. However, there are times when we encounter something interesting or useful that we'd like to remember. That's when we turn to our phones for assistance.
While note apps are useful for text-based information, for visually rich details, we resort to photos. Quite often, we're taking photos not to capture cherished moments with family and friends, but to remember practical things - items like a receipt, a product package, a document, or a car's license plate. It's as if we're using our phones as a tool to extend our limited visual memory.
No dedicated App for the Task
You'd think with so many of us using our phones this way, there'd be a dedicated app to manage these "memory aid" photos, right? Well, not quite. Sure, there are tons of cameras, galleries, and social networks for your "normal" photos, but when it comes to these utility photos, there's no real efficient solution. There have been some attempts to create “photo note” tools, but none have really hit the mark. They often end up being too complex, asking users to categorize photos into folders or tags, which is way more work than just snapping a photo with the default camera. Even the popular "second brain" note apps, as great as they are for typical notes, fall short here. They need manual organization and aren't really built with photos in mind.
So, without a dedicated tool, we end up using our main camera and gallery to remember things, which comes with its own set of issues:
- Clutter: Utility photos get mixed up with your "normal" photos, cluttering the gallery and making it harder to find your regular photos.
- Storage: They take up as much space as a high-quality 12MP photo, even if you're just trying to remember something as simple as a receipt.
- Finding photos: With everything mixed up and no powerful search engine, it's a real pain to find utility photos when you need them.
NeuralBox - Your AI Memory Extension
We believe that with the right tool, we could use our phones to remember stuff a lot more than we currently do. Imagine an app designed from the ground up to let you store anything you find interesting or useful, as much as you'd like, and then easily retrieve it when needed - just like a good external visual memory would.
We've built such an app, NeuralBox, and we're thrilled to present it to you today. Let's dive into its main features:
Designed for loads of Utility Photos
NeuralBox is specifically engineered to store a multitude of utility photos. It allows you to efficiently store as many photos as you want without cluttering your main gallery. With features like two-tap capture, multiple input modes, efficient storage, and hybrid neural search, it's designed to make entering, storing, and retrieving anything you find interesting or useful a breeze. Let’s take a look at each of these features in more detail.
With NeuralBox, remembering something is as simple as two taps. Just tap the lockscreen widget, capture, and you're done. There's no need for folders or tags; everything is organized automatically, making the process quick and hassle-free.
Multiple Input Modes
Besides the two-tap capture, NeuralBox provides several other input modes to suit different situations. For precise document capture, we have a dedicated scanning mode. If you want to import older utility photos from your gallery, you can use the import feature. Additionally, any screenshots taken after the app is installed are automatically imported on each app start, ensuring all your visual memories are conveniently stored in one place.
Just as you don't worry about running out of neurons when trying to remember something, your artificial brain shouldn't be any different. NeuralBox stores photos in an optimized, compressed, and lower resolution format that retains essential information while consuming minimal space. With the optional Plus subscription, you can enable cloud storage, effectively eliminating concerns about storage space.
Hybrid Neural Search
Adding many photos to NeuralBox is quick and straightforward. But how do you find them later without using folders or tags? This is where the magic happens. NeuralBox employs what we call Hybrid Neural Search, designed to help you search and browse your artificial brain as intuitively as possible. This search system comprises multiple components, so let's delve into what happens for each photo added to NeuralBox:
OCR & Keyword search: First we use Optical Character Recognition to extract all text from the photo, store it, and index it to make it searchable later using good old keyword search. This is very useful because in lots of types of utility photos (ex. docs) text is the most important part.
Semantic Image Search & Browsing: We then leverage the latest Neural Networks to generate high-quality image embeddings for the photos, enabling us to perform semantic image search. This is a fascinating aspect of the technology, so let's delve a bit deeper.
In recent years, significant advancements in Artificial Intelligence have led to the popular use of embeddings for semantic text and image search. In simple terms, these image embeddings are a set of numbers generated by a specially trained neural network for a given image. The intriguing part is that images with similar content produce similar sets of numbers, much like how our brain groups related concepts together.
Once these embeddings are generated, we can use them for Image Similarity Browsing. Given an input image and its embedding, we can find similar images by calculating the distance between their embeddings. The "closest" images are the most visually similar, allowing us to "browse our brain" by jumping from one related concept to another, just like how our brain naturally works.
Image Similarity Browsing (using the 'More like this' button) for an image of a bottle of wine displays images that are the closest to the source image. In this case, the first two images are of the same bottle of wine in the same setting, followed by another visually similar bottle in the same setting, and then other wines.
But what about text search? This is where the process becomes even more interesting. While the Image Similarity Browsing feature allows us to find visually similar images, what if we want to search for images semantically using text? This is where OpenAI's model, CLIP, comes into play.
Introduced in early 2021, CLIP bridges the gap between text and images. Its unique feature is the ability to generate both Image AND Text embeddings, while maintaining the principle that "similar embeddings are closer together". This means we can take a text input (like the search term "car"), generate a text embedding for it, and then look up all the images that have image embeddings close to it. The results should be about cars or items semantically related to cars (like tires).
This is how Semantic Image Search works. It mirrors the natural associative processes of our brains, allowing us to search for images using text in a way that feels intuitive and natural.
Semantic Image Search for 'wine' brings up images containing objects closely related to the concept of wine, such as wine bottles, glasses, wine openers, etc.
Hybrid Neural Search: Finally, when you conduct a search, we blend these two methods (Keyword search on OCR text and Semantic Image Search) to help you locate images that either contain the search text within the image or depict any of the objects described in your search term. This hybrid search, combined with semantic image browsing, enables you to swiftly find most images in your NeuralBox, regardless of the quantity.
Hybrid Neural Search for the same concept of 'wine', as above, brings up images that not only contain objects closely related to the concept of wine but also images that include the text 'wine' somewhere within the image.
The Capture Everything Habit
With NeuralBox, you can capture anything and everything, turning your phone into an external memory aid that allows for effortless retrieval later. Once you get the hang of it, you'll find yourself developing a new habit of simply "capturing everything." You might be surprised at how often these random captures come in handy, even days, weeks, or months down the line. Not only that, but this creates a unique visual log of your encounters and thoughts, making it not only useful but also fascinating to look back on. It's a new way to see and navigate your past experiences and insights.
NeuralBox is available in two versions:
Free: Ideal for those wanting to test the waters, this lightweight version utilises local storage (photos are not backed up to our cloud).
Plus: For those seeking the full AI visual memory experience, our Plus version offers 50GB of cloud storage for photo storage and backup, along with data sync across devices. It also features an advanced scanning mode and we'll remove the NeuralBox watermark from your shared images.
Get it in the AppStore
So there you have it! All you've got to do now is head over to the AppStore, download the app, pop that lock screen widget on for speedy captures, and start stuffing your NeuralBox with all the interesting and handy things you come across. Trust us, your actual brain will be pretty grateful.