v1.1.0

FaceDetection v1.1.0

Diff since v1.0.1

Training optimisations (src/Adaboost.jl diff since v1.0.1); bug fix in calculations of weights (src/HaarLikeFeature.jl diff since v1.0.1) and faceness measure (src/Utils.jl diff since v1.0.1)

Details

• Made info/progress optional
  • Add ENV flag for displaying logging information
• Optimised parts of Adaboost algorithm
• Use generics where possible
• Ensure functions use integral arrays where required (it does not make theoretical sense to work with any array)
• Slight bug fix in calculation of weights
• Added example script using FFHQ dataset for faces and THINGS dataset for objects, with around 70–80% accuracy for both
  • Scripts to download (and filter) these datasets
• Convention fixes in code (Blue Code Style + Julia Lang conventions)
• Added interface to give functions a Vector of files as well as the directory itself
• Fixed faceness measure
• Infer feature size if too large for image
• Set minimum compatible Julia version to 1.4

Merged pull requests:

• CompatHelper: bump compat for "ImageIO" to "0.4" by @github-actions in #48
• Added 64-bit x86 to CI by @jakewilliami in #49
• CompatHelper: bump compat for "ImageIO" to "0.5" by @github-actions in #51
• CompatHelper: bump compat for "Images" to "0.24" by @github-actions in #52
• CompatHelper: bump compat for "ColorTypes" to "0.11" by @github-actions in #53
• Merge master into optimisation branch by @jakewilliami in #55
• Optimisation by @jakewilliami in #57
• Improve faceness measure by @jakewilliami in #59
• Added a quick start section to README.md by @jakewilliami in #60
• CompatHelper: add new compat entry for QuartzImageIO at version 0.7, (keep existing compat) by @github-actions in #61
• CompatHelper: add new compat entry for ImageView at version 0.11, (keep existing compat) by @github-actions in #62
• CompatHelper: add new compat entry for Images at version 0.25, (keep existing compat) by @github-actions in #63
• CompatHelper: add new compat entry for FileIO at version 1, (keep existing compat) by @github-actions in #64
• CompatHelper: add new compat entry for Netpbm at version 1, (keep existing compat) by @github-actions in #65
• CompatHelper: add new compat entry for ProgressMeter at version 1, (keep existing compat) by @github-actions in #66
• CompatHelper: add new compat entry for ImageDraw at version 0.2, (keep existing compat) by @github-actions in #67
• CompatHelper: add new compat entry for ColorTypes at version 0.11, (keep existing compat) by @github-actions in #68
• CompatHelper: add new compat entry for ImageMagick at version 1, (keep existing compat) by @github-actions in #69
• CompatHelper: add new compat entry for IntegralArrays at version 0.1, (keep existing compat) by @github-actions in #70
• CompatHelper: add new compat entry for ImageIO at version 0.6, (keep existing compat) by @github-actions in #71
• CompatHelper: add new compat entry for ImageIO at version 0.6, (keep existing compat) by @github-actions in #72

v1.0.1

FaceDetection v1.0.1

Diff since v1.0.0

Bug fix in get_score and area calculations for three_horizontal areas. Bug fix in indexing while creating features (see Caveats in the README).

Merged pull requests:

• Dev (#42) (@jakewilliami)
• Updated examples in README.md with support for serialisation of votes… (#43) (@jakewilliami)
• some changes (#44) (@longemen3000)
• more improvements (#45) (@longemen3000)

v1.0.0

FaceDetection v1.0.0

Diff since v0.2.0

Added sequential processing

v0.2.0

FaceDetection v0.2.0

Diff since v0.1.1

Added sequential processing

Merged pull requests:

• Dev (#35) (@jakewilliami)
• Updated .travis.yml (#36) (@jakewilliami)
• Updated .travis.yml (#37) (@jakewilliami)

v0.1.1

FaceDetection v0.1.1

Diff since v0.1.0

Added scaling images features; switched to sequential processing of images and altered allocating array processing for memory efficiency

Closed issues:

• Use sparse arrays to save on memory (#25)
• Fix speed of sequential processing (#29)

Merged pull requests:

• CompatHelper: bump compat for "Images" to "0.23" (#27) (@github-actions[bot])
• Mem (#28) (@jakewilliami)
• CompatHelper: bump compat for "Netpbm" to "1.0" (#30) (@github-actions[bot])

v0.1.0

FaceDetection v0.1.0

An implementation of Viola Jones' face detection algorithm written purely in Julia.

Closed issues:

• Print messages in src/FDA.jl with colour (#2)
• Ensure $\alpha_t=\log\left(\frac{1}{\frac{\varepsilon_t}{1-\varepsilon_t}}\right)$ (#3)
• Ensure calculations for weighting is true to the original algorithm (#5)
• Adjust getVote function to see at which point the cascade stops, and return that "face-likeness" score (#7)
• Rename function names to be in line with Julia coding convention (#9)
• Refine getIntegralImage as per help from JuliaImages people (#10)
• Fix face detection accuracy (#11)
• Ensure all libraries are needed (#12)
• clean up code after bug fixing (#13)
• Consider changing deep* functions back to Base functions (#14)
• Make files modules and put exports at the start (#15)
• Ensure all print functions are inside of FDA.jl (i.e., the main programme) (#16)
• Ensure getScore and getVote have strict parameter typing (#17)
• Determine minFeatureHeight, maxFeatureHeight, minFeatureWidth, and maxFeatureWidth better (#18)
• Be explicit in which modules the functions come from (#19)
• Statistically analyse the differences in box plots. (#22)
• Get file name in csv (#23)
• Use dict to define feature_types rather than referencing ordered array (#24)