Here we release the TRaffic ANd COngestionS (TRANCOS) dataset, a novel benchmark for (extremely overlapping) vehicle counting in traffic congestion situations. It consists of 1244 images, with a total of 46796 vehicles annotated. All the images have been captured using the publicly available video surveillance cameras of the Dirección General de Tráfico of Spain.

TRANCOS Dataset Sample images with annotations


Best practice: Recommendations on using the dataset

TRANCOS dataset is divided into three dataset: training, validation and test. Any approach reporting results for the TRANCOS bechmark must be trained using any data except the provided test images. Furthermore, the test data must be used strictly for reporting of results alone - it must not be used in any way to train or tune systems, for example by running multiple parameter choices and reporting the best results obtained. For that purpose we recommend to use the training and validation sets which are provided on the same package.

We encourage two types of participation: (i) methods which are trained using only the training+validation set of images provided by TRANCOS; (ii) methods trained using any data except the test data of TRANCOS.


We evaluate vehicle counting performance using the GAME metric detailed in our IbPRIA 2015 paper.

In the following table we show the latest result reported in the TRANCOS benchmark. All methods are ranked based on the GAME metric for L=3.

Rank Method GAME(L=0) GAME(L=1) GAME(L=2) GAME(L=3)
1 Hydra CCNN (ECCV2016) 10.99 13.75 16.69 19.32
2 IbPRIA 2015 (Lemptisky+SIFT) 13.76 16.72 20.72 24.36
3 IbPRIA 2015 (Fiaschi+RGB Norm+Filters) 17.68 19.97 23.54 25.84
4 IbPRIA 2015 (HOG-2) 13.29 18.05 23.65 28.41

Add Your Results

To incorporate your results to this table, please, send us ( the following information:

  • The GAME (for L=0,1,2,3) of your method.
  • A link to the scientific publication where you describe the technical solution implemented.

Database Rights

The database has been made publicly available for scientific research purposes. All the images have been captured using the publicly available cameras of the Dirección General de Tráfico of Spain. Use of these images must respect Dirección General de Tráfico terms of use.


This work is supported by projects CCG2014/EXP-054, IPT-2012-0808-370000, TEC2013-45183-R and SPIP2014-1468 of the Dirección General de Tráfico of the Government of Spain.


If you make use of this data and software, please cite the following reference in any publications:

  Title                    = {Extremely Overlapping Vehicle Counting},
  Author                   = {Ricardo Guerrero-Gómez-Olmedo, Beatriz Torre-Jiménez, Roberto López-Sastre, 
  							  Saturnino Maldonado Bascón, and Daniel Oñoro-Rubio},
  Booktitle                = {Iberian Conference on Pattern Recognition and Image Analysis (IbPRIA)},
  Year                     = {2015}