Comments (4)
Ah I see that VGG16 in Caffe will take variable sized images, and thus you end up with variable size feature maps. That seems mildly odd though -- feeding in the same image at two different scales would presumably yield a different embedding. Any thoughts on whether it makes sense to scale everything before running through crow
?
from crow.
We reported some comparisons on this topic in our paper, but the takeaway is that, at least on benchmark datasets, keeping the full resolution is worth ~5-10% mAP compared to 224x224 and only ~2-4% mAP compared to 586x586. In general, the higher resolution, the better mAP, with diminishing returns. Depending on the demands of your application, it may make sense to resize to a fixed size (e.g. especially if it enables efficient batching on GPU hardware or the like). The version here keeps the full resolution for the best possible retrieval performance with an implicit assumption that all photos you might extract features for are roughly the same size.
from crow.
Excellent thanks. 586x586 sounds reasonable for this application in that case -- the original images are huge but sortof variably sized. As an anecdote, scaling them to a fixed size appeared to give a nontrivial bump in performance, compared to passing them in at a variety of sizes.
from crow.
“The version here keeps the full resolution for the best possible retrieval performance”, but the input size is 10 * 3 * 224 * 224 in VGG_ILSVRC_16_pool5.prototxt. Will it resize the img into 224 * 224 automatically when running?
The problem I met is the poor performance when applying to imgs whose "height > width".
@pumpikano
from crow.
Related Issues (8)
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 crow.