3ed397bacd
When compiled with cargo build --features=analytics and enabled via moonfire-nvr run --object-detection, this runs object detection on every sub stream frame through an Edge TPU (a Coral USB accelerator) and logs the result. This is a very small step toward a working system. It doesn't actually record the result in the database or send it out on the live stream yet. It doesn't support running object detection at a lower frame rate than the sub streams come in at either. To address those problems, I need to do some refactoring. Currently moonfire_db::writer::Writer::Write is the only place that knows the duration of the frame it's about to flush, before it gets added to the index or sent out on the live stream. I don't want to do the detection from there; I'd prefer the moonfire_nvr crate. So I either need to introduce an analytics callback or move a bunch of that logic to the other crate. Once I do that, I need to add database support (although I have some experiments for that in moonfire-playground) and API support, then some kind of useful frontend. Note edgetpu.tflite is taken from the Apache 2.0-licensed https://github.com/google-coral/edgetpu, test_data/mobilenet_ssd_v2_coco_quant_postprocess_edgetpu.tflite. The following page says it's fine to include Apache 2.0 stuff in GPLv3 projects: https://www.apache.org/licenses/GPL-compatibility.html
Introduction
Moonfire NVR is an open-source security camera network video recorder, started
by Scott Lamb <slamb@slamb.org>. It saves H.264-over-RTSP streams from
IP cameras to disk into a hybrid format: video frames in a directory on
spinning disk, other data in a SQLite3 database on flash. It can construct
.mp4 files for arbitrary time ranges on-the-fly. It does not decode,
analyze, or re-encode video frames, so it requires little CPU. It handles six
1080p/30fps streams on a Raspberry Pi
2, using
less than 10% of the machine's total CPU.
So far, the web interface is basic: a filterable list of video segments, with support for trimming them to arbitrary time ranges. No scrub bar yet. There's also no support for motion detection, no https/SSL/TLS support (you'll need a proxy server, as described here), and only a console-based (rather than web-based) configuration UI.
This is version 0.1, the initial release. Until version 1.0, there will be no compatibility guarantees: configuration and storage formats may change from version to version. There is an upgrade procedure but it is not for the faint of heart.
I hope to add features such as salient motion detection. It's way too early to make promises, but it seems possible to build a full-featured hobbyist-oriented multi-camera NVR that requires nothing but a cheap machine with a big hard drive. I welcome help; see Getting help and getting involved below. There are many exciting techniques we could use to make this possible:
- avoiding CPU-intensive H.264 encoding in favor of simply continuing to use the camera's already-encoded video streams. Cheap IP cameras these days provide pre-encoded H.264 streams in both "main" (full-sized) and "sub" (lower resolution, compression quality, and/or frame rate) varieties. The "sub" stream is more suitable for fast computer vision work as well as remote/mobile streaming. Disk space these days is quite cheap (with 3 TB drives costing about $100), so we can afford to keep many camera-months of both streams on disk.
- decoding and analyzing only select "key" video frames (see wikipedia).
- off-loading expensive work to a GPU. Even the Raspberry Pi has a surprisingly powerful GPU.
- using HTTP Live Streaming rather than requiring custom browser plug-ins.
- taking advantage of cameras' built-in motion detection. This is the most obvious way to reduce motion detection CPU. It's a last resort because these cheap cameras' proprietary algorithms are awful compared to those described on changedetection.net. Cameras have high false-positive and false-negative rates, are hard to experiment with (as opposed to rerunning against saved video files), and don't provide any information beyond if motion exceeded the threshold or not.
Documentation
- License — GPLv3
- Building and installing
- UI Development
- Troubleshooting
- Wiki has notes on several camera models.
Getting help and getting involved
Please email the moonfire-nvr-users mailing list with questions, or just to say you love/hate the software and why. You can also file bugs and feature requests on the github issue tracker.
I'd welcome help with testing, development (in Rust, JavaScript, and HTML), user interface/graphic design, and documentation. Please email the mailing list if interested. Pull requests are welcome, but I encourage you to discuss large changes on the mailing list or in a github issue first to save effort.