Tesla released a new version of their FSD Beta 10.4 on November 7th, according to Tesla Raj on Twitter, the updates include:
- Improving handling by adjusting regenerative braking while leaving the navigation route safely
- Improving processing and detection of high-speed objects to enable faster acceleration on highways
- Improving vehicle speed on narrow roads with obstacles
- Improving static obstacle recognition by 1.5% precision and 7% recall with hyperparameter tuning and improved sampling strategy
- Improving pedestrian, bicycle, and motorcycle recognition rates by 35% in precision, 20% in recall, and 1.1% in control accuracy
- Improving detection system for specialty vehicles with a 5.8% increase in passing rate
- Taking pictures using external ADAS cameras when the top of the car screen is pressed for “video recording” and sharing them with Tesla engineering team to help improve FSD.
One of the most noticeable improvements is the automatic recognition and avoidance of a police car parked on the side of the road while following another lane with turn signal activated, as shown in a video shared by a Twitter user recently.
It seems that the FSD Beta has made substantial progress in both safety and accuracy, making true autonomous driving by Tesla more feasible. However, Tesla has not announced a timeline for deploying FSD in China, and we hope to have access to the system as soon as possible.
🔗Source: teslarati
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.