What is it?

The Shashin Go Concept 1 is an Open Source Hardware design for AI data capture. It is a 5mp camera based on the Sony Spresense. It is a DIY wearable. It is a programmable camera platform. You decide what you want to do with it.

Why did you make it?

I have always enjoyed making things, and I have been frustrated by portable and wearable cameras that required an app to install, an account to create, and things that are not compatible with any other camera software I use. I wanted a stack that I could own top to bottom, code, and mix and match parts that I want to use. I experimented with a number of different lower cost camera based AI platforms, however, the camera performance was disappointing. The Sony Spresense is a solid design, high quality camera, and full of interesting features for a DIY Maker.

My goal was to have something I could easily 3D print a case, and create a wearable design I could use at a tradeshow. I brought this to CES2024 and walked the floor the entire day for 3 days, and captured stills of everything. I can then go back and review the data collected from the show, catalog ideas, update my notes, or even use this data for training some AI model. For now, I am mostly interested in data capture for memory sake.

My second goal was to have full control of the design. Aside from decisions by Sony, I can choose how this thing works. And so can you. You can add your own sensors. The case is modular, so you can add things inside. If you don't like the button I picked, choose a different one. Want a different LED, add it. The Sony Spresense main board and extension board are nice, and the docs are good enough. You are limited by your skills and time only.

My third goal was to make this easy. I wanted to walk a show floor for 8+ hours and not worry about batteries. Done. Plugin an external USB pack and it seems to run for close to 14 hours, depending on your USB battery pack (Not included). How do you get the pictures off? Just pop out the SDCard (not included). Put it into your PC. Totally easy. No stuff to install.

What makes it special?

1. Open Design: I will release the 3D files for the case, which I got from an open source design. The software will be on Github. I modified the existing Camera tester to work as an autofiring camera, with settings tuned for my use case. You can modify the firmware to suit your needs easily.
2. Own your Stack principle: Gone are the days where you need to cede control to a monopolistic company. They own your data. They force you to have an account. They make you receive marketing information. They sell your data. They mine your data. They don't support what they sell after a finite number of years. If I made it, I'll continue to support it. But you own it. It's yours and it will continue to work.
3. Control the quality. I have control over the camera output. If I want to change the output format to JPEG or PNG, I can do that. If I want some video stream captured, I can do that. Control is important. Everything doesn't have to be curated by a big tech company.
4. Invest in longevity. If you build it, you can keep it working indefinitely. Gone are the days of replacing your stuff every few years because the vendor won't support it.

The ugly:

• It's too big. I plan to do another run with a smaller package that fits onto a tag.
• It's expensive. I don't like the price. But this is basically what the components cost in this low volume. I won't make money on this first cut design. By paying the cost, you are supporting future designs which will aim to reduce the footprint.
• It lacks wireless. One of the frustrating aspects of the Sony Spresnse. It doesn't shop with any wireless interfaces. Odd choice, and the boards they do support as add ons are expensive and hard to find. I personally am ok without wireless in this first design. But you might mind.
• It looks gaudy. Yes, that was the point. A HAL9000 type LED on front. A big colorful button. As I said, it's a DIY thing. You can customize.
• It lacks a built in battery. Power management is tricky. I'll add it in the next version.
• The camera doesn't have autofocus. Yes, it's a fixed focus camera. That's how they work. You can manually spin the focus, but it's somewhat hard to do. A future design might have a fancier lens but this one is actually pretty good for the size.
• No USB mass storage. I'd love to plug this in and have it show up as a USB mass storage device. Not in this version.
• Low light support is lacking. Camera performance isn't great in low lighting. No surprise. You can tweak the settings to optimize low light performance.
• It isn't a sports / action cam. I don't know that this is bad thing, but be aware that there are no active moving parts on the camera so the lens isn't refocusing as objects move. To change the ISO speed, you may want to just set it to automatic. For ML applications, you generally don't need the highest resolution images, and perhaps prefer a black and white image.
• No RTC. Not having a persistent time keeper is an issue. This will be used in a future iteration, but I am aware of the need. I'd like to improve the metadata encoded into the images and the image naming scheme. I have provided a reasonable workaround which is to organize the photos into sets which can be sorted using the time relative to the local epoch.

Software

The original software used in this DIY Camera is based on the original Sony Spresense "Test" sample. Instead of using the serial port for input, the camera has a single operating mode: shooting pictures or not. I've tweaked the camera modes a bit to suit the use case of walking around with a wearable. This sample is based on the Arduino ecosystem and you can tweak it further from our git repo. The big useful feature I added is to organize data into "sets". Each press of camera trigger opens a new set1 ... set5 ... setN. Photos get organized by incremented value so pictures can easily be sorted by time.

I will be posting an additional example that is a "person detector". You can use this to train your own model. This uses the Nettx RTOS and Spresense SDK, which is a more complex and powerful ecosystem. I expect this to be the ecosystem that any power users will select.

A Reward for your support

Anyone who purchases this first concept will be entitled to a 30% discount on any future Shashin Go camera concepts I make. The next version I build, I hope to go a step further in two directions. You will just need to indicate the invoice on any future purchases and I can offer the discount at that time in the future.

• Focus: A standalone, small, portable package with a battery, and built in gesture sensor to control the camera trigger and some other unique features to save certain moments that have just occurred "save the last minute of data". Has WiFi for sharing photos.
• Focus: A wearable building block for makers to use specific AI features built into the camera to trigger things in an "if this then that" sort of design. An example would be: the camera sees faces it recognizes, sends an audio alert to a bluetooth audio platform that is listening for notifications from the camera. Or, simply, notify you of danger in the environment. Many use cases.