![](bg7.png)
(a) Frame corresponding to the activation maps.
(b) Activation maps produced by the first convolution layer.
Figure 4: Activation maps from Model 1 trained for 200 epochs.
using only videos as input. As seen in Section 3, it
is possible to observe that the third proposed model
trained with 350 epochs obtained a lower NRMSE
when compared to the others, including the single
channel as reference model. Specifically in this mo-
del, it can be seen that M-CNNs provide significant
improvements in their use in autonomous vehicle ap-
plications. The performance of the best model was
approximately 7% better than the reference model.
Furthermore, the fifth model presents results simi-
lar to the others, showing that it was capable of main-
taining robustness even when receiving an input with
reduced dimensionality.
Future works may include the addition of explicit
space-time information during the training stage, cre-
ating new versions of datasets and training models,
exploring how M-CNNs architecture respond to this
information. Also, new datasets and more complex
situations must be tested to validate the approach.
REFERENCES
Abadi, M., Agarwal, A., Barham, P., Brevdo, E., Chen, Z.,
Citro, C., Corrado, G. S., Davis, A., Dean, J., De-
vin, M., Ghemawat, S., Goodfellow, I., Harp, A., Ir-
ving, G., Isard, M., Jia, Y., Jozefowicz, R., Kaiser,
L., Kudlur, M., Levenberg, J., Man
´
e, D., Monga, R.,
Moore, S., Murray, D., Olah, C., Schuster, M., Shlens,
J., Steiner, B., Sutskever, I., Talwar, K., Tucker, P.,
Vanhoucke, V., Vasudevan, V., Vi
´
egas, F., Vinyals, O.,
Warden, P., Wattenberg, M., Wicke, M., Yu, Y., and
Zheng, X. (2015). TensorFlow: Large-scale machine
learning on heterogeneous systems. Software availa-
ble from tensorflow.org.
Baccouche, M., Mamalet, F., Wolf, C., Garcia, C., and Bas-
kurt, A. (2011). Sequential deep learning for human
action recognition. In International Workshop on Hu-
man Behavior Understanding, pages 29–39. Springer.
Bojarski, M., Testa, D. D., Dworakowski, D., Firner, B.,
Flepp, B., Goyal, P., Jackel, L. D., Monfort, M., Mul-
ler, U., Zhang, J., Zhang, X., Zhao, J., and Zieba,
K. (2016). End to end learning for self-driving cars.
CoRR, abs/1604.07316.
Bojarski, M., Yeres, P., Choromanska, A., Choroman-
ski, K., Firner, B., Jackel, L. D., and Muller, U.
(2017). Explaining how a deep neural network trai-
ned with end-to-end learning steers a car. CoRR,
abs/1704.07911.
Chen, C., Seff, A., Kornhauser, A. L., and Xiao, J. (2015).
Deepdriving: Learning affordance for direct percep-
tion in autonomous driving. CoRR, abs/1505.00256.
Autonomous Vehicle Steering Wheel Estimation from a Video using Multichannel Convolutional Neural Networks
523