Comentários e feedback de alunos de Visual Perception for Self-Driving Cars da instituição Universidade de Toronto

I know this is a tough topic, but I was expecting more in-depth and practical coverage of the object-detection and segmentation CNNs. Week 1&2 gave a good overview of visual perception and feature detection. Week 3-6 were pretty shallow. The final project was reasonably well-put, but the outputs of the object detector and the segmentation CNNs were just given for use and to me personally those are the most interesting aspects of the autonomous vehicle vision system.

While the lectures were fairly well done they in no way prepared you for the projects. Way too much time was spent deciphering exactly what was to be done in the project sections. Very disappointing was the complete lack of any support on the forum - none - zero. A little feedback and support here would haver made all the difference. Absolutely do NOT recommend this course and will not be taking the following on. Just not worth the time. I am better off learning on my own.

Content is great but lack of instructor support makes the course hard to understand.

Lots of errors in assignments. I had to read forums for almost every graded assignment, that's disappointing.

Some programming assignments need to be provided with more guidance and clarification.

need more example in lesson for programming tasks and for equations

Great delivery! Mr. Waslander explains the most complex technical stack in such a easy-to-understand way. It is most beneficial for students who have no experience in the visual perception field but want to have a general grasp of the overall workflow. Much learnt for me personally.

This course is good organised and contain the general representation of the perception task. But I'd like to notice several cons:

1 The material about neural networks doesn't have any practical exercises with NN training or inference.

2 The second assessment gave me a lot of nervous ours/days/months. It hasn't any expected results during the long code and compare only the final result at the submitting. But the most strange thing is that during this assessment student is supposed to do matrix equations that are not desired in lecture or supplementary material

3 As one of the basic neural networks VGG is considered. But in practice the VGG is very time consuming to use. This is important fact, as in self-driving vehicles we need to stay in real-time. VGG is definitely bad choice for that. It would be nice to add some overview of the fast NNs for object detection.

1.The professor's video only has very rough and basic introduction to the topics, while the first programming homework needs much more self_learning on many aspects. What disappointed me is that I often cannot get enough help from Coursera's resources, including forums.

2. I started learn this course aiming to apply what I have learned of CNN to some practical applications, but this course neither had in-depth introduction to the build of self-driving car CNN architecture nor relevant homework to practice it. It is the most disappointing part to me.

3. The final homework does not include most of the knowledge taught in this session but only focus on the semantic segmentation. I hope the final project could be more comprehensive..

Overall, I graded this course 3 stars, for the missing 2 stars due to very shallow introduction of CNN and superficial video contents.

The lecture is very clear and concise but not cover too many technical details. The program assignment especially the one of week 2 is very struggling and took me almost two months to succeed by continuing the course subscriptions. Most struggling time is spent in trying different OpenCV models with tunning parameters, and debug instead of learning key skills. I don't if this is workable for the real self-driving practice. I hope that more instructions could have been given for the program assignment.

Like the previous two course I found this one well structured and presented. Basically my comments from course 1 and 2 still hold. I found the coding assignment for week 2 rather challenging but with the help of the discussion forum there should be no problem to pass it. In contrast the final coding project was less difficult. I really loved the content of the final assignment as it provided a detailed look under the hood of a perception stack guiding you through the various stages. The multiples pictures generated as part of your code are a great help to understand the various aspects.

Very informative, structured, and resourceful course. It provides a very strong theoretical background on Visual Perception. The practicals also are well designed with sets of instructions to understand and work through them (although it is essential to have a good python and openCV background). The lecturer is very knowledgeable about the topic. Now, I have a good direction for my research work (thanks to the different areas of active explorations introduced by the lecturer).

I do not understand why this course just have 4.3 ratting. Personally I think this course is very very helpful. It provides many practical advice and makes feel that I have got a up-to-date understanding of this fiels. There is no doubt that this is one of the best courses on Coursera.

Though not dive into training neural net. But for me who have taken deep learning specialization, I fully respect this and find it amazing that this course introduces quite a lot of the application of deep learning output and provides programming exercises on them, which is great.

This course is excellent! It covers a broad range of basics of computer vision to in depth image detection and object collision estimation. I'd recommend this to anyone looking for a thorough introduction to visual perception for self-driving cars.

Overall a great learning experience. Some "heavy" topics like state-of-the-art object detection using CovNets are not thoroughly treated (that would take an entire course), but references are provided. The best part is the programming assignments.

The course has proved to another milestone in furthering my understanding of robotics, computer vision, machine learning and autonomous driving vehicles.

Really really great course. I would like to work with Prof.Waslander at any project. I will advise this course to anyone interested. Thanks Coursera!

I am really surprised at the depth of topics discussed. I believe i spent around 5-8 hours researching topics on ANN and Machine learning.

superb, the assignment was quite tough but the overall experience was amazing. thanks to instructors, TAs, Coursera, and fellow learners!

Wonderful course. I knew little things about gps and imu sensors. Camera is a complete new concept to me. Thank you.

although I have been working with object detection and image segmentation things but still alot of learning

Many thanks for this amazing course!!!! was very hard to me but I have learned a lot!!! Thanks!!!

give a detailed and intuitive presentation on visual perception process for self driving cars.

It is an amazing course. Really good information and projects related with Visual Perception