|
|
Vehicle Counting in Still Image
Vágner, Filip ; Juránek, Roman (referee) ; Špaňhel, Jakub (advisor)
The goal of this work is to compare models of convolutional neural networks designed to count vehicles in a static image using density estimation with a focus on different sizes of objects in the scene. A total of four models were evaluated - Scale Pyramid Network, Scale-adaptive CNN, Multi-scale fusion network and CASA-Crowd. The evaluation was done on three data sets - TRANCOS, CARPK, PUCPR+. Scale Pyramid Network achieved the best results. The model reached 5.44 in the Mean Absolute Error metric and 9.95 in the GAME(3) metric on TRANCOS dataset.
|
|
|
Vehicle Counting in Still Image
Jelínek, Zdeněk ; Juránek, Roman (referee) ; Špaňhel, Jakub (advisor)
The main goal of this thesis was to compare different approaches to vehicle counting by density estimation. Four convolutional neural networks were tested - Counting CNN, Hydra CNN, Perspective-Aware CNN and Multi-column CNN. The evaluation of these models was done on three different datasets. The Perspective-aware CNN has achieved the most accurate results across all datasets. This model has reached 2.86 Mean Absolute Error on the PUCPR+ dataset, proving that it is the most suitable for the vehicle counting problem.
|
|
|
Counting Vehicles in Image and Video
Gabzdyl, Dominik ; Herout, Adam (referee) ; Špaňhel, Jakub (advisor)
Analýza silničního provozu je stále náročnou úlohou. V průběhu této úlohy se vyskytují mnohá úskalí, která je třeba brát na vědomí. Například malé rozlišení obrazu, vysoký počet překrývajících se objektů, úhel kamery, rozmazání objektů v důsledku jejich pohybu nebo povětrnostní podmínky. Tato práce adresuje tato úskalí použitím konvolučních neuronových sítí. V této práci představuji novou architektu založenou na principu počítání regresí (Counting by Regression). Navržená architektura je inspirována některými state-of-the-art architekturami a vylepšuje přesnost na různých datasetech. Například na velmi malém PUCPR+ datasetu byla odmocnina ze střední kvadratické chyby (RMSE) snížena z 34.46 na 6.99 vozidel (měřeno na test setu). Dosažené výsledky ukázaly, že je zde stále prostor ke zlepšení a možný další výzkum v oblasti počítání regresí.
|
|
|
Occupancy Estimation of a Parking Lot from Images
Dubovec, Pavol ; Špaňhel, Jakub (referee) ; Herout, Adam (advisor)
When determining the number of vehicles in the pictures of carparks that do not have the appropriate parameters needed for processing, the problem of vehicle counting can be quite complex. The aim of this work is to create an application that detects the number of vehicles in the selected photo, regardless of selected carpark view. This detection will be performed using machine learning, based on a model, created by training on trained data, which consists of photographs of parking lots from different perspectives and positions. The problem was solved in an unconventional way, by splitting the pictures with the parking lot into several areas of interest (zones) and creating anotations from these areas, using the created application specialized for this task. The images are then formatted to the same size. These prepared cutouts are then loaded to the Keras API, which is used to train the model. The aim was to create a model that would be versatile enough to determine the number of vehicles in a photograph in any environment (time, weather, weather conditions) and in the shortest possible time. Currently, the model can predict the correct number of vehicles in the cutout on test data with an accuracy of 87% and with a first order error of 95%. This work focuses on solving this problem in real time. It is a classification into 7 classes (0-6 vehicles). This solution could be interesting especially for static cameras in atypical places (eg side view), or it is important for them to capture certain areas.
|
|
|
Vehicle Counting in Still Image
Hladiš, Martin ; Juránek, Roman (referee) ; Špaňhel, Jakub (advisor)
The goal of this thesis is to compare different models of convolutional neural networks, which use the principle of using density estimation to count the number of vehicles in a still image. The tested models were -- Counting CNN, Scale-adaptive CNN, Multi-Scale Fusion Net a Multi-scale CNN. Their estimation capability was tested using these datasets -- TRANCOS, CARPK, PUCPR+. The most accurate results were achieved by the Multi-Scale Fusion Net model. Its estimation accuracy using the dataset TRANCOS in the Mean Absolute Error metric achieved value of 8.05.
|
|
|
Vehicle detection in images
Pálka, Zbyněk ; Přinosil, Jiří (referee) ; Krajsa, Ondřej (advisor)
This thesis dissert on traffic monitoring. There are couple of different methods of background extraction and four methods vehicle detection described here. Furthermore there is one method that describes vehicle counting. All of these methods was realized in Matlab where was created graphical user interface. One whole chapter is dedicated to process of practical realization. All methods are compared by set of testing videos. These videos are resulting in statistics which diagnoses about efficiency of single one method.
|
|
|
Vehicle Counting in Still Image
Vágner, Filip ; Juránek, Roman (referee) ; Špaňhel, Jakub (advisor)
The goal of this work is to compare models of convolutional neural networks designed to count vehicles in a static image using density estimation with a focus on different sizes of objects in the scene. A total of four models were evaluated - Scale Pyramid Network, Scale-adaptive CNN, Multi-scale fusion network and CASA-Crowd. The evaluation was done on three data sets - TRANCOS, CARPK, PUCPR+. Scale Pyramid Network achieved the best results. The model reached 5.44 in the Mean Absolute Error metric and 9.95 in the GAME(3) metric on TRANCOS dataset.
|
|
|
Vehicle Counting in Still Image
Hladiš, Martin ; Juránek, Roman (referee) ; Špaňhel, Jakub (advisor)
The goal of this thesis is to compare different models of convolutional neural networks, which use the principle of using density estimation to count the number of vehicles in a still image. The tested models were -- Counting CNN, Scale-adaptive CNN, Multi-Scale Fusion Net a Multi-scale CNN. Their estimation capability was tested using these datasets -- TRANCOS, CARPK, PUCPR+. The most accurate results were achieved by the Multi-Scale Fusion Net model. Its estimation accuracy using the dataset TRANCOS in the Mean Absolute Error metric achieved value of 8.05.
|
|
|
Occupancy Estimation of a Parking Lot from Images
Dubovec, Pavol ; Špaňhel, Jakub (referee) ; Herout, Adam (advisor)
When determining the number of vehicles in the pictures of carparks that do not have the appropriate parameters needed for processing, the problem of vehicle counting can be quite complex. The aim of this work is to create an application that detects the number of vehicles in the selected photo, regardless of selected carpark view. This detection will be performed using machine learning, based on a model, created by training on trained data, which consists of photographs of parking lots from different perspectives and positions. The problem was solved in an unconventional way, by splitting the pictures with the parking lot into several areas of interest (zones) and creating anotations from these areas, using the created application specialized for this task. The images are then formatted to the same size. These prepared cutouts are then loaded to the Keras API, which is used to train the model. The aim was to create a model that would be versatile enough to determine the number of vehicles in a photograph in any environment (time, weather, weather conditions) and in the shortest possible time. Currently, the model can predict the correct number of vehicles in the cutout on test data with an accuracy of 87% and with a first order error of 95%. This work focuses on solving this problem in real time. It is a classification into 7 classes (0-6 vehicles). This solution could be interesting especially for static cameras in atypical places (eg side view), or it is important for them to capture certain areas.
|
|
|
Vehicle Counting in Still Image
Jelínek, Zdeněk ; Juránek, Roman (referee) ; Špaňhel, Jakub (advisor)
The main goal of this thesis was to compare different approaches to vehicle counting by density estimation. Four convolutional neural networks were tested - Counting CNN, Hydra CNN, Perspective-Aware CNN and Multi-column CNN. The evaluation of these models was done on three different datasets. The Perspective-aware CNN has achieved the most accurate results across all datasets. This model has reached 2.86 Mean Absolute Error on the PUCPR+ dataset, proving that it is the most suitable for the vehicle counting problem.
|
guest ::
