AI News, Using neural networks to detect car crashes in dashcam footage

Using neural networks to detect car crashes in dashcam footage

That’s why when you’re asked to prove yourself human through those CAPTCHA tests, you’re always asked a ridiculously simple question, e.g., whether an image contains a road sign or not, or selecting a subset of images that contain food (see Moravec’s Paradox).

In this post, I will describe how, as a Fellow for Insight Data Science, I built a classification machine learning algorithm (Crash Catcher!) that employs a hierarchical recurrent neural network to isolate specific, relevant content from millions of hours of video.

For businesses that may have millions of hours of video to sift through (for instance, an auto insurance company), the tool I created is extremely useful to automatically extract important and relevant content.

In the ideal world, the perfect data set for this particular problem would be a large repository of videos with thousands upon thousands of examples of both car crashes and non-crashes.

Each video would also contain clear metadata and be consistent in terms of video (e.g., location of camera in car, video quality and duration) and content (e.g., type of car crash, whether it be “head-on” or “t-bone”).

Their videos were short, four second-long clips with approximately 600 videos of accidents and about a thousand videos without accidents (just normal, boring driving scenes).

If I were to train an algorithm on the 439 negative examples and 36 positive examples in my data set, the resulting model could easily predict that there were no crashes and still have 92.5% accuracy.

Each image can be thought of as a two-dimensional array of pixels (originally with a size of 1280x720), where each pixel has information about the red, green, and blue (RGB) color levels which creates a 3-D shape.

Video datasets are particularly challenging because of their structure — while each frame in the video might be understood using standard image recognition models, understanding overall context is more difficult.

This approach allowed me to train a model to understand the flow of features and objects within a single video, and translate that into patterns that differentiate videos with crashes from videos without crashes.

The first neural network analyzes the time-dependent sequence of the images within each video, tracing objects or features as they move or change throughout the clip (e.g., car headlights or car bumpers).

The second recurrent neural network takes the patterns and features encoded by the first neural network and learns patterns to discern which videos contain accidents and which do not.

I tuned the model’s hyperparameters (e.g., number of layers of neurons, number of videos loaded into memory at a time, loss function, number of epochs) to optimize for accuracy, where the training and validation sets iterated through different options of these hyperparameters.

You may be wondering about that other 20% of the data I haven’t mentioned — these 52 videos were my holdout test set to analyze the final performance of the model after tuning the model.

Reducing false negatives (crashes incorrectly identified as normal driving scenes) is key, as these are the cases that are most important to companies sifting through large quantities of video, even if there are occasional videos without crashes predicted to have crashes (false positives).

To train this algorithm to accurately predict a wider variety of situations (e.g., classifying head-on collision vs rear-end collision vs car-truck collisions), more data is necessary.

Video of Uber Self-Driving Car’s Fatal Crash Raises More Questions

Information continues to emerge about the automated vehicle in Uber’s fleet that fatally struck a pedestrian the night of 19 March 2018 in Tempe, Ariz.

While Tempe police chief Sylvia Moir cautiously speculated that neither a human nor an automated driver could have avoided the crash based on video of the incident, she refused to rule out charges for the backup driver.

The lidar system used on Uber’s automated vehicle to detect pedestrians works just as well in complete darkness as in broad daylight, yet initial reports said that neither the automated driving system nor the test driver applied the brakes until after impact.

There are three ways the lidar might have failed: It never detected the pedestrian, it detected her but misclassified her as a benign object such as a bush or an exhaust cloud, or it classified her correctlybut failed to accurately predict her movements, perhaps assuming she would stop in the turn lanes.

Even the information we have now may not be accurate—while police and most media outlets reported that the speed limit was 35miles per hour, a Google Street View image from July 2017 clearly shows a 45 mph speed limit sign a mile before the crash site.

Monitoring driver attention is particularly important in highly automated vehicles, as studies consistently show that off-road glances increase dramatically when the throttle, brake, and steering are simultaneously under computer control.

Modeling Vehicle Collision Angle in Traffic Crashes Based on Three-Dimensional Laser Scanning Data

A vehicular collision is complicated and traditional methods of accident investigation, such as inspection of the vehicle’s body, trace fixing, and photographing the vehicle, do not provide accurate information regarding collision angle.

Road traffic accident reproduction technology refers to reproducing the form of the accident process accurately, according to the visible traces, in combination with the description of the accident from witnesses following traffic accidents.

[5] developed a vehicle continuous collision accident reappearance system by studying the mechanism of vehicle continuous collision, using the trajectory prediction iterative algorithm, refactoring localization algorithm of contact position, inverse association with the accident, etc., which provides a good reference value for collision angle.

in digital images of traffic accidents, there is significant difficulty in representing traffic accidents based on image data, so the image analysis method has been studied to ascertain if it could be a suitable tool in accident reappearance studies.

[7] proposes a method based on a multivariate Monte Carlo solution, setting speed, braking, steering, and many other factors as the variables for input, and using the Monte Carlo method to test the form of the movement before the accident, thus innovatively applying the method to traffic accident reappearance.

[10] built a vehicle collision dynamics model that describes the changes in vehicle body structure before and after a collision, and the study produced accelerated speed and displacement data under different collision angles and car body stiffness values.

[11] studied the critical collision angle of light-duty trucks and established a death time series model of fatal collision angle using traffic accident death statistics.

Car collision angle analysis is an integral part of the study of vehicle collisions, which could make accident reconstruction more accurate than early car crash studies that relied on crash tests alone.

As for 3D reconstruction scanning equipment, the major producers emphasize different indexes, including ranging precision, ranging scope, data sampling rate, the distance between the minimum point, modeling point positioning accuracy, size of the laser spot, scanning field, laser level, and laser wavelength.

[14] built a 3D scene model reconstruction system called AEST (Automatic Environmental Sensor for Telepresence), which integrates many types of 3D reconstruction algorithms including triangulation of three-dimensional point cloud data, data registration, data fusion, etc.

However, the specifics of gaining deformation information based on the scanning point cloud and then using the deformation characteristics to obtain the vehicle collision angle presents a new problem worthy of study.

2018 Audi A8 - intelligent Drive

Pro Sense Side 00:00 Dynamic All-Wheel Steering 00:35 Rim Protection 01:05 Remote Garage Pilot 01:59 Remote Parking Pilot 03:03 Manoeuvre Assist 04:15 ...

Traffic car accident reconstruction and investigation Solution

- The new innovative Solution for traffic car accident ..

Mythbusters - Car crash force

Rolling a Reliant Robin - Top Gear - BBC

Jeremy takes the extreme sport of Reliant Robin rolling to the streets of Barnsley, aided by a string of celebrities who just happen to be on hand to help keep the ...

The Universal Hot vs Crazy Matrix - a Man's Guide to Women

"Everything a young man needs to know about women".

Boomerang Trick Shots | Dude Perfect

Time to take boomerangs to the next level! ▻ Click HERE to subscribe to Dude Perfect! ▻ Click HERE to watch our most recent ..

DUAL CAMERA CAR DASH CAM - 2.7" 1080P Full HD DVR - Dome G30B

Gear Best Link for this Dash Camera: Gear Best Car DVR Category: ..

Why all world maps are wrong

Making accurate world maps is mathematically impossible. Subscribe to our channel! Interact with projections: ..

PupSaver 35 mph Crash Test Video

PupSaver has been deemed crash test-worthy by a sled test conducted at MGA Research Corporation by our safety design engineer, using the following criteria: ...