edgefirst-codec

JPEG and PNG decoding straight into pre-allocated tensors — no per-frame allocations, with optional hardware acceleration on Linux.

PyPI License

Part of the EdgeFirst HAL

edgefirst-codec is one of five Python packages built from the EdgeFirst Hardware Abstraction Layer.

The EdgeFirstAI/hal repository is the home for all of them — source, issue tracker, architecture documentation and release notes.

Package Provides
edgefirst-tensor Zero-copy tensor allocation and host/GPU/CUDA mapping
edgefirst-codec JPEG and PNG decoding directly into pre-allocated tensors (this package)
edgefirst-image GPU-accelerated colour conversion, resize, letterbox, tiling and drawing
edgefirst-decoder YOLO and ModelPack output decoding
edgefirst-tracker ByteTrack multi-object tracking

Installation

pip install edgefirst-codec

Requires Python 3.8 or newer; edgefirst-tensor and NumPy are installed automatically. Wheels are published for Linux (x86_64, aarch64), macOS (arm64), and Windows (x86_64), and are self-contained — there is no system JPEG library to install.

Packages install under the PEP 420 edgefirst.* namespace, so the import is edgefirst.codec.

Quick start

For maximum performance, decode straight into a tensor allocated by edgefirst.image's ImageProcessor.create_image() — DMA/PBO-backed and GPU-pitch-aligned — then hand it to ImageProcessor.convert() for colour conversion and resize. decode_file_into / decode_into are free functions rather than Tensor methods precisely so they can take a tensor from another edgefirst.* package:

import numpy as np
from edgefirst.codec import Tensor, decode_file_into
from edgefirst.image import Flip, ImageProcessor, PixelFormat, Rotation

processor = ImageProcessor()

# peek_image_info_file reads the header only — no pixels are decoded.
info = Tensor.peek_image_info_file("frame.jpg")
print(info.width, info.height, info.format)  # e.g. 1280 720 PixelFormat.Nv16

# Allocate once, outside the loop. The decoder reconfigures the tensor's
# dimensions and format within this allocation, so one tensor sized for the
# largest expected frame can receive smaller images without reallocating.
src = processor.create_image(
    info.width, info.height, PixelFormat.Nv12, "uint8", "readwrite"
)
dst = processor.create_image(640, 640, PixelFormat.Rgb, "uint8", "readwrite")

info = decode_file_into(src, "frame.jpg")  # or decode_into(src, jpeg_bytes)
# convert() performs colour conversion (native → RGB) and resize; the codec
# reports EXIF orientation in `info` but does not apply it, so pass it on.
rotation = Rotation.degrees_clockwise(info.rotation_degrees)
flip = Flip.Horizontal if info.flip_horizontal else Flip.NoFlip
processor.convert(src, dst, rotation, flip)

with dst.map() as view:
    data = np.frombuffer(view, dtype=np.uint8)

Same-package pipelines that never leave edgefirst.codec can use the equivalent Tensor.decode_image_file() method instead — tensor.decode_image_file("frame.jpg") — but decode_image_file is a method, and its self must literally be an edgefirst.codec.Tensor; a DMA/PBO-backed destination from edgefirst.image.ImageProcessor.create_image() is a different package's type and can never be that self, so it must go through the free functions decode_into / decode_file_into instead.

Images decode in their native pixel format and are never colour-converted, rotated or resized. A colour JPEG lands on the NV format matching its own chroma sampling (4:2:0 → Nv12, 4:2:2 → Nv16, 4:4:4 → Nv24), greyscale on Grey, and PNG on Rgb / Rgba / Grey. Nothing is resampled on the way out.

EXIF orientation is reported, never applied: info.rotation_degrees and info.flip_horizontal carry the transform your pipeline should apply downstream, and the reported dimensions are unrotated.

Tuning the decode

from edgefirst.codec import DctMethod, PixelFormat, set_dct_method, set_output_format

set_dct_method(DctMethod.Fast)  # faster IDCT, small bounded accuracy cost
set_output_format(PixelFormat.Rgb)  # fused colour conversion inside the decode

Both settings are thread-local — apply them on every thread that decodes. set_output_format fuses colour conversion into the decode's MCU write stage, which is a pure-CPU single-pass path; pass None to restore native output.

PixelFormat and the other value types (TensorMemory, Region, the colour axis enums) are accepted from any edgefirst.* package, not just this one — they compare and hash equal across packages by value, so ==, dict keys and set membership all work regardless of which package's copy you pass. Tensors, Decoder and ProtoData cross packages too, through the capsule protocols. Importing PixelFormat from edgefirst.codec when calling into this package is still good style for readability, not a requirement. See the Interoperability section below for the one thing that does not cross: isinstance against a concrete class.

What this package provides

API Purpose
Tensor.peek_image_info() / peek_image_info_file() Read dimensions, format and EXIF orientation from the header alone
Tensor.decode_image() / decode_image_file() Decode into a pre-allocated edgefirst.codec.Tensor
decode_into() / decode_file_into() Decode into a tensor from any edgefirst.* package (e.g. edgefirst.image.ImageProcessor.create_image())
ImageInfo Decoded dimensions, native format, row stride, EXIF orientation
set_dct_method() / DctMethod IDCT accuracy/speed selection
set_output_format() Fused Rgb / Nv12 decode output
is_v4l2_available() Whether a V4L2 hardware JPEG decoder is present

Hardware acceleration

On Linux the decoder transparently tries hardware backends before the software path and falls back without any API change:

  • V4L2 mem2mem — SoC JPEG blocks such as the i.MX mxc-jpeg. Discovery is capability-based, with no hardcoded device node. Opt out with EDGEFIRST_DISABLE_V4L2=1.
  • nvJPEG — CUDA GPU decode on NVIDIA platforms such as Jetson Orin. Loaded via dlopen, so there is no link-time CUDA dependency. Opt in with EDGEFIRST_ENABLE_NVJPEG=1.

Footprint

Adding edgefirst-codec to a project costs roughly 1.5 MB of downloads (this package plus edgefirst-tensor), excluding NumPy. That is roughly an order of magnitude smaller than Pillow or OpenCV, and a few times larger than a bare libjpeg-turbo binding — while bundling PNG, EXIF and the hardware backends, with no system libraries to install. See the Rust crate README for the comparison table.

Supported inputs

JPEG decoding covers baseline DCT, 8-bit precision, 1 or 3 components. Progressive, lossless, hierarchical and arithmetic-coded JPEG, CMYK/YCCK and non-8-bit precision are rejected with a typed error rather than mis-decoded. PNG goes through zune-png and supports 8-bit and 16-bit Luma / LumaA / RGB / RGBA. The Rust crate README documents the full matrix.

Interoperability

decode_into / decode_file_into accept a tensor from any edgefirst.* package because each extension module registers its own Tensor type object (PyO3 issue #1444 — isinstance across packages is always False, even for two objects wrapping the same Rust type). Acceptance goes through the __edgefirst_tensor__ capsule protocol every Tensor implements:

# CORRECT — works regardless of which edgefirst.* package produced obj
if hasattr(obj, "__edgefirst_tensor__"):
    ...

# WRONG — always False for a tensor from a sibling package
if isinstance(obj, edgefirst.image.Tensor):
    ...

edgefirst.tensor.EdgeFirstTensorExportable (re-exported here as edgefirst.codec.EdgeFirstTensorExportable) is a typing.Protocol you can annotate a cross-package parameter with. See crates/python-common/INTEROP.md for the full protocol.

Versioning and changelog

All four edgefirst-* packages are versioned and released together with the HAL itself, so a given version number refers to the same source tree in every language. Because of that there is no per-package changelog: release notes for every version live in the single CHANGELOG.md in the hal repository, which follows Keep a Changelog and Semantic Versioning.

Links

License

Apache-2.0

Project links

Data verified by PyPI on Sep 4, 2026
Data provided by the project maintainers, verified at the time the release was uploaded to PyPI.

Key dates

PyPI data
Data sourced directly from PyPI's database.
  • Released:
Newer release available (0.30.0)

1 maintainer

PyPI data
Data sourced directly from PyPI's database.
Avatar for au-zone from gravatar.com au-zone

Credits

PyPI data
Data sourced directly from PyPI's database.

Author: Au-Zone Technologies

GitHub Statistics

Data verified by PyPI on Sep 4, 2026
The GitHub source repository was provided by the project maintainers and verified by PyPI at the time of upload. Stars, forks, and open issues/PRs are derived from that repository and have not been independently verified.

License expression

Apache-2.0
View SPDX License List

Requires

Python >=3.8

Tags

jpeg png image-decoding zero-copy edge-ai

Classifiers

Report project as malware

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distributions

No source distribution files available for this release.See tutorial on generating distribution archives.

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

edgefirst_codec-0.29.3-cp311-abi3-win_amd64.whl (651.9 kB view details)

Uploaded CPython 3.11+Windows x86-64

edgefirst_codec-0.29.3-cp311-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (806.3 kB view details)

Uploaded CPython 3.11+manylinux: glibc 2.17+ x86-64

edgefirst_codec-0.29.3-cp311-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (782.4 kB view details)

Uploaded CPython 3.11+manylinux: glibc 2.17+ ARM64

edgefirst_codec-0.29.3-cp311-abi3-macosx_11_0_arm64.whl (705.0 kB view details)

Uploaded CPython 3.11+macOS 11.0+ ARM64

edgefirst_codec-0.29.3-cp38-abi3-win_amd64.whl (657.5 kB view details)

Uploaded CPython 3.8+Windows x86-64

edgefirst_codec-0.29.3-cp38-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (810.6 kB view details)

Uploaded CPython 3.8+manylinux: glibc 2.17+ x86-64

edgefirst_codec-0.29.3-cp38-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (787.2 kB view details)

Uploaded CPython 3.8+manylinux: glibc 2.17+ ARM64

edgefirst_codec-0.29.3-cp38-abi3-macosx_11_0_arm64.whl (712.9 kB view details)

Uploaded CPython 3.8+macOS 11.0+ ARM64

File details

Details for the file edgefirst_codec-0.29.3-cp311-abi3-win_amd64.whl.

File metadata

File hashes

Hashes for edgefirst_codec-0.29.3-cp311-abi3-win_amd64.whl
Algorithm Hash digest
SHA256 bc683e37c0d0c7280252b6d6dd1cf2026ed3ee4f2579ad54e23f686b107e9fe5
MD5 f1da4e22745f27bdea5273e6f6ca4248
BLAKE2b-256 28ca947138b4c8179c56a18ddac975963fe8681e3b9347c4c3f1d070cc8c2220

See more details on using hashes here.

Provenance

The following attestation bundles were made for edgefirst_codec-0.29.3-cp311-abi3-win_amd64.whl:

Publisher: release.yml on EdgeFirstAI/hal

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file edgefirst_codec-0.29.3-cp311-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.

File metadata

File hashes

Hashes for edgefirst_codec-0.29.3-cp311-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
Algorithm Hash digest
SHA256 f478e0ca19ca467770c72520eed10dd2e15100627e7429a9f6f39664048226d4
MD5 eaf54d8b734435ad88928f5e3852bc54
BLAKE2b-256 01265d208517c2e46df12dcee6832d39f0d45124d09235c4af7f5b8b29c95767

See more details on using hashes here.

Provenance

The following attestation bundles were made for edgefirst_codec-0.29.3-cp311-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl:

Publisher: release.yml on EdgeFirstAI/hal

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file edgefirst_codec-0.29.3-cp311-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl.

File metadata

File hashes

Hashes for edgefirst_codec-0.29.3-cp311-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
Algorithm Hash digest
SHA256 7b2f522e5684b879f0224fcecf9345b6bcdebb9c6e9bd3e04f85450bda023098
MD5 ebf69ce37d70bfe5dcaa1a84fe7c0d9a
BLAKE2b-256 891229951beb91274a6a289b08dd4c3f8688da561d3fe6efcb3beb99c223c32f

See more details on using hashes here.

Provenance

The following attestation bundles were made for edgefirst_codec-0.29.3-cp311-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl:

Publisher: release.yml on EdgeFirstAI/hal

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file edgefirst_codec-0.29.3-cp311-abi3-macosx_11_0_arm64.whl.

File metadata

File hashes

Hashes for edgefirst_codec-0.29.3-cp311-abi3-macosx_11_0_arm64.whl
Algorithm Hash digest
SHA256 4bd0a53a58bf31ecd67b903368a820fc9e2940acf7fb18a5188bd246ed254b81
MD5 e30f414ef2851d6a643c30737cd82357
BLAKE2b-256 7cbe854654906a99b0a83e816abe3050b6b2c664c23bf1e6a8936f676d520331

See more details on using hashes here.

Provenance

The following attestation bundles were made for edgefirst_codec-0.29.3-cp311-abi3-macosx_11_0_arm64.whl:

Publisher: release.yml on EdgeFirstAI/hal

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file edgefirst_codec-0.29.3-cp38-abi3-win_amd64.whl.

File metadata

File hashes

Hashes for edgefirst_codec-0.29.3-cp38-abi3-win_amd64.whl
Algorithm Hash digest
SHA256 fbd63619277de6d972914a46dc6a51f6a62a53c6dd149e52a68f43226783261b
MD5 659f339884540e1d47f0d4ba2746c08d
BLAKE2b-256 aee171278a4e837e49c119297d59d7384b982ff27b8607ebd81e16d1dd49fdfb

See more details on using hashes here.

Provenance

The following attestation bundles were made for edgefirst_codec-0.29.3-cp38-abi3-win_amd64.whl:

Publisher: release.yml on EdgeFirstAI/hal

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file edgefirst_codec-0.29.3-cp38-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.

File metadata

File hashes

Hashes for edgefirst_codec-0.29.3-cp38-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
Algorithm Hash digest
SHA256 78dada85277895b06f208a6be71b06cb2cd62d7bd1a3d859dc58c2c74f98263a
MD5 eb7ada6c1c4657997e4555efc1313385
BLAKE2b-256 f27d562812ba1e245e3d9d9c965874392fda7454c2017e642d5015ebf8f27037

See more details on using hashes here.

Provenance

The following attestation bundles were made for edgefirst_codec-0.29.3-cp38-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl:

Publisher: release.yml on EdgeFirstAI/hal

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file edgefirst_codec-0.29.3-cp38-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl.

File metadata

File hashes

Hashes for edgefirst_codec-0.29.3-cp38-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
Algorithm Hash digest
SHA256 c7a1579a3fdec6693ed84212715ffe5a7c3af77e0fe9c1f6878cdcd823b5cc35
MD5 02e4b8133f0207853edc8e223e890633
BLAKE2b-256 cc7c4d014cefa1d7bf34ae7715d70c75f9e643be8d79a61f917bcd648592a598

See more details on using hashes here.

Provenance

The following attestation bundles were made for edgefirst_codec-0.29.3-cp38-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl:

Publisher: release.yml on EdgeFirstAI/hal

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file edgefirst_codec-0.29.3-cp38-abi3-macosx_11_0_arm64.whl.

File metadata

File hashes

Hashes for edgefirst_codec-0.29.3-cp38-abi3-macosx_11_0_arm64.whl
Algorithm Hash digest
SHA256 4b4107b8763145fb54c18f4f2a22d8d3e294a4bec27d7da89664d94fe1e86a2c
MD5 70f2f67ee2903bf3c94a9509876dad82
BLAKE2b-256 4a215700457b626911a90cffa436a96146d20a1fd32e66154142b3d505eafcab

See more details on using hashes here.

Provenance

The following attestation bundles were made for edgefirst_codec-0.29.3-cp38-abi3-macosx_11_0_arm64.whl:

Publisher: release.yml on EdgeFirstAI/hal

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Release history Release notifications | RSS feed

0.30.0

8 files

0.29.4

8 files

This release

0.29.3 This release

8 files

Pre-release

0.29.0rc1

1 file

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page