v1.2.0

v1.2.0 is a foundational, very large release that builds out most of the classic OpenMV imaging API: the drawing methods, binary / morphology operations, keypoint descriptors (FAST / FREAK / LBP), color conversions and image.statistics, BMP / PPM / JPEG file I/O, the lcd, mjpeg, gif, and mlx thermal modules, the OV7725 camera, and many sensor controls. Some core sensor behavior changed — read the breaking changes below.

Highlights

  • Drawing API — draw_line() / draw_rectangle() / draw_circle() / draw_string() / draw_cross() / draw_keypoints().

  • Binary / morphology — binary(), invert(), and/or/xor/..., erode() / dilate() / morph(), negate() / difference().

  • Keypoints — FAST / FREAK / LBP descriptors with match_descriptor() / save_descriptor() / load_descriptor().

  • New modules — lcd, mjpeg, gif (color), and mlx thermal.

  • OV7725 camera support; image.statistics; BMP/PPM/JPEG file I/O.

  • Breaking: sensor.reset(), sensor.snapshot(), and sensor.set_pixformat() behavior changed — see the breaking changes.

New features

  • Drawing — added image.draw_line() / draw_rectangle() / draw_circle() / draw_string() / draw_cross() / draw_keypoints() with keyword arguments (color, thickness, …).

  • Binary / morphology — added image.binary(), invert(), and() / nand() / or() / nor() / xor() / xnor(), erode() / dilate() / morph(), negate() / difference(), pixels(), centroid(), orientation_radians() / orientation_degrees(), and width() / height() / format().

  • Keypoints / descriptors — FAST keypoint extraction via find_keypoints(), a unified match_descriptor() and save_descriptor() / load_descriptor() for FREAK / LBP, plus a profile-face Haar/LBP cascade.

  • Color / statistics — image.statistics and color-conversion functions (rgb_to_lab / lab_to_rgb / rgb_to_grayscale / grayscale_to_rgb).

  • File I/O — BMP / PPM (P2/P3/P5/P6) / JPEG load and save (including grayscale JPEG and 4:2:0 / 4:2:2 chroma subsampling).

  • Modules — new lcd (LCD shield), mjpeg (video recording), color gif (with Gif.loop()), and mlx thermal-camera modules, each with example scripts.

  • Sensor — added sensor.get_id() (+ OV9650/OV2640/OV7725 PID constants), sensor.set_special_effect() (SDE_* constants), sensor.set_image_filter() (FILTER_BW / FILTER_SKIN), the HQVGA framesize, and image.set_pixel(x, y, …).

  • WiFi — added the ATWINC1500 WiFi driver and OpenMV 1/2 BSP groundwork, plus a BLE example.

Other changes and improvements

  • The framebuffer is now JPEG-compressed before streaming to OpenMV IDE (lower bandwidth); JPEG compression is done in place; OV7725 H/V zoom enables arbitrary scaling; improved (faster) Haar face detection that runs at QVGA.

Bug fixes

Imaging:

  • Fixed the integral-image computation, median() filter corruption, the centroid computation, RGB skin / black-and-white filters on RGB input, and faster grayscale BMP/PPM read/write.

Sensor and system:

  • Fixed the pyb.Servo timers (Servo now works), HQVGA resolution, SD-card init reliability, re-initializing storage after a soft-reset, and re-running the REPL when no script is present.

Hardware and board support

  • OV7725 camera sensor.

  • LCD shield (lcd module), mlx thermal camera, and ATWINC1500 WiFi groundwork.

Breaking API changes

User-visible API breaks between v1.1.0 and v1.2.0. Scope: Python C-modules in modules/ and Python libraries in scripts/libraries/.

All three breaking changes are behavior changes to core sensor functions (same API, different behavior) — re-check affected scripts. Each commit hash links to its diff on GitHub. (The vast majority of this release is new API built out within the v1.1→v1.2 range and is therefore additive, not breaking.)

sensor.reset() no longer applies a default config (behavior)

sensor.reset() no longer applies a built-in default configuration (pixformat / framesize / framerate / gain / contrast / brightness / saturation). Scripts that relied on reset() leaving a usable default must now explicitly call sensor.set_pixformat() and sensor.set_framesize() (and any other settings) after reset().

Commits: d1e782df3

sensor.snapshot() raises on timeout (behavior)

sensor.snapshot() now raises RuntimeError("Sensor Timeout!!") when the sensor times out instead of hanging indefinitely. Wrap snapshot() in try / except RuntimeError if you need to tolerate timeouts.

Commits: e3f41d674

sensor.set_pixformat() raises on unsupported format (behavior)

sensor.set_pixformat() now raises an exception when given an unsupported pixel format instead of returning False. Code that checked the False return must catch the exception instead.

Commits: e0c5822c2

Migration checklist

For a clean port to v1.2.0 the typical work is:

  1. Explicitly set sensor.set_pixformat() / sensor.set_framesize() after sensor.reset() (the reset change).

  2. Wrap sensor.snapshot() in try / except RuntimeError if you need to tolerate sensor timeouts (the snapshot change).

  3. Catch the exception from sensor.set_pixformat() instead of checking for False (the set_pixformat change).

All other scripts run unchanged.