Comments (7)
Namely after "smart" feature selection
from facedetection.jl.
Done in commit e911698.
from facedetection.jl.
#25 addresses this broader issue of memory efficiency
from facedetection.jl.
Commit 61b0382 addresses this
from facedetection.jl.
Points to consider and possibly implement:
- GPU processing
- Use
ArrayFire.jl
'sload_image
- Use
CUDA.jl
- Use
- Use 4 x ones or -ones using
Int8
(ask @dmipeck) -
pmap
-
Mmap
- Use
SharedArray
s
from facedetection.jl.
Commit 70d7d74 addresses this
from facedetection.jl.
Some overview thus-far.
These benchmarking results are from tests since we changed the algorithm to run sequentially.
Commit | Benchmark Time of Tests (seconds) | % Time Improvement Since Previous Listed Commit | Number of Allocations | Memory Allocation | % Memory Improvement Since Previous Listed Commit |
---|---|---|---|---|---|
a4689195 | 30.689 | — a | 371318354 | 6.38 GiB | — a |
da9c833e | 7.768 | 74.69 | 104464112 | 2.51 GiB | 60.66 |
b3aec6b8 | 5.025 | 35.31 | 28589987 | 713.89 MiB | 71.56 |
3e9be4ad | 5.242 | -4.32 b | 46688538 | 990.05 MiB | -38.68 b |
a I did not benchmark prior to this, though it probably wouldn't be hard to checkout
and rewrite some tests with an older version.
b In this commit, I had to change the output of the get_vote
function from Int8
to Float64
for correctness. As a result, this is why we have a decrease in benchmarks since the previous listed commit.
NB—: time improvement since last commit can be calculated very easily:
julia> improvement(a, b) = ((a - b) / a) * 100
improvement (generic function with 1 method)
julia> improvement(30.689, 7.768)
74.68799895728111
That is to say, there was a 74.7% improvement between times 30.689 s and 7.768 s.
from facedetection.jl.
Related Issues (20)
- Ensure all `print` functions are inside of FDA.jl (i.e., the main programme) HOT 1
- Ensure `getScore` and `getVote` have strict parameter typing HOT 1
- Determine minFeatureHeight, maxFeatureHeight, minFeatureWidth, and maxFeatureWidth better
- Be explicit in which modules the functions come from
- Read pretrained data from file HOT 7
- Scale classifiers (to ensure that classifiers work on images different sizes to training data) HOT 1
- Statistically analyse the differences in box plots.
- Get file name in csv
- Use dict to define feature_types rather than referencing ordered array
- Use sparse arrays to save on memory HOT 3
- Fix speed of sequential processing
- Add docstrings
- Fix discrepancy with pythonic results HOT 2
- TagBot trigger issue HOT 2
- Failing on 32-bit machines
- Zygote integeration
- Ensure correctness of faceness measure HOT 2
- Ensure IntegralArray adheres to AbstractArray interface HOT 1
- Support rotated Haar features
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from facedetection.jl.