Collection of subarasii and reproducible SBIR works.
Criteria: works must have codes available, and the reproducible results demonstrate state-of-the-art.
This collection is inspired by the image denoising collection. It makes me have an idea, why don't I summarize the SBIR works just like he did.
- GF-HoG
[Web][Web] [Code] [PDF]- Gradient field descriptor for sketch based retrieval and localization (ICIP, 2010), Rui Hu et al.
- A Bag-of-Regions Approach to Sketch Based Image Retrieval (ICIP, 2011), Rui Hu et al.
- A Performance Evaluation of Gradient Field HOG Descriptor for Sketch Based Image Retrieval (COMPUT VIS IMAGE UND, 2013), Rui Hu et al.
- Color GF-HoG [Web] [Code] [PDF]
- Scalable Sketch-based Image Retrieval using Color Gradient Features (ICIP, 2015), Tu Bui et al.
- SHOG [Code(msvs)] [Code(origin)]
- Sketch-based image retrieval: Benchmark and bag-of-features descriptors (IEEE T VIS COMPUT GR, 2011), Eitz M et al.
- A descriptor for large scale image retrieval based on sketched feature lines (SBM, 2009), Eitz M et al.
- How do humans sketch objects? (TOG, 2012), Eitz M et al.
-
Classification
- Sketch-a-Net that Beats Humans (BMVC, 2015), Qian Y et al. [PDF] [trained model] & Sketch-a-Net: A Deep Neural Network that Beats Humans (IJCV, 2016), Qian Y et al. [Code(Tensorflow)] [Code(Tensorflow)(modified net)] [Code(Matlab)] [Code(Pytorch)] [PDF]
-
Triplet loss CNN
- The Sketchy Database: Learning to Retrieve Badly Drawn Bunnies (TOG, 2016), Sangkloy P et al. [Web] [Code(Caffe)] [PDF]
- Compact Descriptors for Sketch-based Image Retrieval using a Triplet loss Convolutional Neural Network (CVIU, 2017), Tu Bui et al. [Web] [Code(Caffe)] [Code(Pytorch)] [PDF]
- Sketch Me That Shoe1 (CVPR, 2016), Yu Q et al. [Web] [Code(Caffe)] [Code(Tensorflow)] [PDF] [Dataset]
- Deep Spatial-Semantic Attention for Fine-grained Sketch-based Image Retrieval1 (ICCV, 2017), Song J et al. [Web] [Code(Tensorflow)] [PDF] [Dataset]
-
DPM
- Fine-grained sketch-based image retrieval by matching DPM1 (unpublished), Yi L et al. [Web] [Code(Matlab)] [PDF]
1: Fine-grained retrieval
- Multi-column weight sharing
- Deep Sketch Hashing: Fast Free-hand Sketch-Based Image Retrieval (CVPR, 2017), Li L et al. [Web] [Code(Caffe)] [PDF]
- Flickr15K
[Web] [Image(source and groundtruth)] [Image(resized)] [Sketch][Image(source and groundtruth)] [Image(resized)] [Sketch] [groundtruth] [PDF]- A Performance Evaluation of Gradient Field HOG Descriptor for Sketch Based Image Retrieval (CVIU 2013), Rui Hu et al.
- 14,660 images labelled into 33 categories based on shape only
- 10 non-expert sketchers(5 males, 5 females)
- 330 free-hand sketches
- Flickr25K [Web] [Image] [PDF]
- Compact Descriptors for Sketch-based Image Retrieval using a Triplet loss Convolutional Neural Network (CVIU 2017), Tu Bui et al.
- 25,000 images labelled into 250 categories
- images only
- can be used associated with TU-Berlin, shown as follows
- TU-Berlin [Web] [Sketch] [PDF]
- How Do Humans Sketch Objects? (Siggraph 2012), Eitz M et al.
- 20,000 unique sketches evenly distributed over 250 object categories
- sketch only
- Image100k [Web] [Image] [Benchmark/Sketch] [PDF]
- Sketch-Based Image Retrieval: Benchmark and Bag-of-Features Descriptors (IEEE T VIS COMPUT GR 2012), Eitz M et al.
- [Benchmark] has 1,240 images labelled into 31 categories and 31 corresponding query sketches for testing
- [Image] has 101,240 images for training
- 28 participants(23 males, 5 females)
- GOLD [Web] [Image(resized)] [PDF]
- Sketch-Based Image Retrieval by Salient Contour Reinforcement (IEEE T MULTIMEDIA 2016), Zhang Y et al.
- contains more than 22,000 Flickr Crawled images together with their Geotags
- extend image set, image only
- Sketchy [Web] [Image&Sketch] [PDF]
- The Sketchy Database: Learning to Retrieve Badly Drawn Bunnies (TOG, 2016), Sangkloy P et al.
- sampled from 125 categories and acquire 75,471 sketches of 12,500 objects
- a benchmark contained
- SketchyScene [Web] [PDF]
- SketchyScene: Richly-Annotated Scene Sketches (ECCV, 2018), Changqing Z et al.