Where Is OpenCV 5?
OpenCV, the widely used open-source computer vision library, is seeking $500,000 via Indiegogo to fund its next major release, OpenCV 5, after losing some core team members and facing chronic funding gaps. Commenters see the amount as small relative to the library’s impact and criticize how large companies rely on it without meaningful financial support, prompting broader debate over sustainable business models, licensing, and corporate responsibility in open source. Many also argue that despite the rise of deep learning frameworks like PyTorch and TensorFlow, OpenCV remains crucial for classical vision tasks, preprocessing, and deployment on low-power devices.
Crowdfunding OpenCV 5 and Project Status
- OpenCV 5 is being crowdfunded via Indiegogo with a target of $500k; progress is in the single‑digit percent range despite front-page exposure.
- The campaign aims to: make 5 a major community-driven release, prove crowdfunding as an alternative to relying on large corporate donors, and establish recurring community-based funding.
- The project reports having lost several core team members due to “global conflicts and instability,” slowing development and issue handling.
Relevance in the Deep Learning Era
- Many argue OpenCV remains highly relevant:
- Provides classical computer vision (edge detection, SLAM, calibration, geometric transforms) where deep learning is overkill, too slow, data-hungry, or unsuitable.
- Widely used for image/video I/O, preprocessing, and augmentation even in TF/PyTorch/ONNX pipelines.
- Essential on low-power/embedded devices where typical NN models are still too heavy or slow.
- Some counter that modern DL frameworks plus small or quantized models can solve most vision tasks “better,” but others note practical deployment limits on microcontrollers and real-time systems.
Corporate Support, Licensing, and Sustainability
- Strong sentiment that widely-used OSS like OpenCV are underfunded while large companies profit from them.
- Explanations offered: tragedy of the commons, corporate preference for paid “enterprise” wrappers with support and accountability, and sensitivity around licenses like GPL or “pay-if-big” models.
- Proposed remedies include: copyleft licensing, revenue/size-based licenses, or explicit “pay to play” relationships, but many note these are often rejected or blacklisted by big companies.
Features, Design Choices, and Pain Points
- Some question funding a deep neural networks module, arguing interop with TF/PyTorch is enough; others say a native C++ DNN stack is strategically important and OpenCV already has a DNN module.
- Complaints about complexity, C++ “dependency nightmare,” unstable APIs, and outdated choices like default BGR color order; others praise continuous performance improvements from v3 to v4.
- Tooling and docs, especially for mobile/Android builds and size-optimized binaries, are seen as difficult; nonetheless OpenCV is described as “indispensable” on platforms lacking strong native vision libraries.