.. _changelog_ide_v2_0_0: v2.0.0 ====== v2.0.0 brings machine-learning support to OpenMV IDE: a new **CNN Network Library** tool for deploying neural-network models to your cam, along with ImageReader video export, a much faster serial terminal, and bundled firmware 3.0.0. This release stays on the Qt Creator 4.0.2 base. There are no breaking changes for existing projects, though Windows installs no longer auto-install the bundled ``pybcdc`` serial driver. .. contents:: On this page :backlinks: none :local: :depth: 1 Highlights ---------- - **CNN Network Library** tool under ``Machine Vision Tools`` copies a bundled or custom ``.network`` model onto the connected OpenMV Cam (or saves it locally when no cam is connected), and ships with ``cifar10``, ``cifar10_fast``, ``lenet``, and ``smile`` models. - **ImageReader export** lets ``Convert Video`` / ``Save Video`` output an OpenMV ``.bin`` file for on-cam playback. - **Faster serial terminal** sends pasted/run code in 63-byte chunks instead of one byte at a time. - **Machine Learning examples** added under ``25-Machine-Learning``, plus new AprilTag and I2C Lidar examples. - **Bundled firmware 3.0.0** for the OPENMV2/3/4 boards, with smaller M7/H7 binaries. New features ------------ - Added the **CNN Network Library** tool under ``Machine Vision Tools``. It copies a bundled or custom neural-network ``.network`` model onto the connected OpenMV Cam, or saves it locally when no cam is connected. The tool ships with ``cifar10``, ``cifar10_fast``, ``lenet``, and ``smile`` models (`5e8c7ce92 `__, `a27537ff1 `__, `f75ef2bcd `__). - The ``Convert Video`` / ``Save Video`` tools can now output an OpenMV **ImageReader** ``.bin`` file, converting a video (or set of images) into a packed ``OMV IMG STR`` stream of RGB565/grayscale frames for playback on the cam (`b7d12cd32 `__). - Added new bundled examples: a ``25-Machine-Learning`` folder with CNN inference scripts (``cifar10``, ``lenet``, and smile-detection, plus center-window and whole-window search variants), a new ``26-April-Tags`` folder (with the AprilTag examples moved out of ``16-Codes``) that adds max-resolution and small-tag detection examples, and an I2C Lidar Lite V3 example under ``24-External-Sensors`` (`56abdf508 `__). Other changes and improvements ------------------------------ - Pasting and running code in the OpenMV serial terminal is much faster: bytes are now sent in (``TABOO_PACKET_SIZE`` - 1)-sized chunks (63 bytes) instead of one byte at a time (`bf93f3856 `__). - The **CNN Network Library** now remembers separate last-used paths for cam-connected versus no-cam saves and stays enabled even when no cam is connected. The bootloader "reconnect your cam" prompt now tells you to hit cancel to skip to DFU reprogramming (`a27537ff1 `__). Bug fixes --------- - Fixed invisible (same-color) menu-bar text on dark and custom themes by inverting the menu text color when the menu background is dark and applying it to ``ButtonText`` as well (`dc7af7b54 `__). - Fixed the **AprilTag generator** so generated tag images are rendered on a proper black border and background and scaled to fit, instead of being placed on an unpainted canvas (`a27537ff1 `__). Platform and tool support ------------------------- - **Qt Creator base:** 4.0.2. - **Bundled OpenMV firmware:** updated to 3.0.0 for the OPENMV2/OPENMV3/OPENMV4 boards, with notably smaller M7/H7 firmware binaries (`8da10eb4b `__). - **Board names:** renamed to the consumer naming scheme: OpenMV Cam M4 (STM32F427), OpenMV Cam M7 (STM32F765/67), and OpenMV Cam H7 (STM32H743) (`8f29f7ca6 `__). - The Windows installer no longer installs the ``pybcdc`` serial driver (`4b27cdb80 `__). Breaking changes ---------------- None for existing projects, scripts, or saved settings. Note that on Windows the bundled ``pybcdc`` serial driver is no longer auto-installed by the installer (`4b27cdb80 `__); this only affects the optional legacy serial driver and not normal cam connectivity.