Patent Drafting Analysis of Tesla’s Enhanced Object Detection for Autonomous Vehicles | US 12,198,396 B2
Patent Drafting Analysis of Tesla's Enhanced Object Detection for Autonomous Vehicles | US 12,198,396 B2
A structural and strategic analysis of Tesla's field-of-view crop-and-downsample object detection patent, examining claim architecture, drafting quality, critical gaps, and prosecution positioning across method, system, and CRM claim types.
Spec Words
5,800
Across 6 sections
Draft now ↗
Total Claims
16
3 independent · 13 dependent
Draft now ↗
Figure Sheets
6
System architecture, method flow, CNN block diagrams
Draft now ↗
Published by PatSnap Insights Team · · 12 min read Verified by PatSnap Eureka Data
Overview
Structural Overview
The detailed description dominates the specification at approximately 67% of total words (~3,900 words), with the claims section representing a solid 27% (~1,560 words across 16 claims), reflecting a well-proportioned technical disclosure. The claim architecture spans 3 independent claims — one method (Claim 1), one system (Claim 8), and one CRM (Claim 15) — with 13 dependent claims providing a 4.33:1 dependent-to-independent ratio. Six drawing sheets cover system architecture, method flowcharts, real-world cropped image examples, and dual CNN pipeline block diagrams, providing good but not exhaustive figure coverage for the disclosed embodiments.
Section Word Distribution
↗ Click bars to exploreFigure Inventory — 6 Sheets
| Figure | Description | Role |
|---|---|---|
| FIG. 1 | System architecture overview showing Image Processing Network (102), Image Database (108), Heuristics Database (110), and Client Device(s) (112) and their interconnections.Search in Eureka ↗ | System architecture |
| FIG. 2 | Flowchart of the object detection method showing six sequential steps: receive image, determine FOV, crop FOV, downsample remainder, send batched inputs to detector, and combine outputs.Search in Eureka ↗ | Flow diagram |
| FIG. 3 | Example road scene image (300) illustrating detected objects surrounded by bounding boxes at the edge of the cropped field of view frame.Search in Eureka ↗ | Claim support |
| FIG. 4 | Schematic representation of the full end-to-end method showing raw image input, priority FOV cropping with horizon label, downsampled remainder, detector block (212), combined outputs, and duplicate removal.Search in Eureka ↗ | System architecture |
| FIG. 5A | Block diagram of first CNN pipeline variant showing input image (502) processed by Conv. Net. A (504A) to extract image features (506) and vanishing line information (508), image portion (510) processed by Conv. Net. B (504B) to produce image portion features (512), combined image features (514) fed to Conv. Net. C (504C) for output (516).Search in Eureka ↗ | System architecture |
| FIG. 5B | Block diagram of second CNN pipeline variant showing downsampled input image (522) processed by Conv. Net. A (504A), with full resolution image portion (530) processed by Conv. Net. B (504B), yielding combined image features (534) for output (516) via Conv. Net. C (504C).Search in Eureka ↗ | System architecture |
Analysis powered by PatSnap Eureka. Patent text and figures publicly available from USPTO. Draft a Similar Patent
Claims
Claim Architecture Analysis
The patent contains 3 independent claims — Claim 1 (method), Claim 8 (system/apparatus), and Claim 15 (non-transitory computer storage media/CRM) — providing tripartite enforcement coverage across all three standard claim types. The dependent-to-independent ratio of 4.33:1 is below the typical 6–8:1 norm for software/AI-driven computer vision patents, suggesting limited fallback depth for prosecution. A notable strategic observation is that Claims 1, 8, and 15 are substantively parallel in structure, with their respective dependent claims (2–7 for method, 9–14 for system, 16 for CRM) mirroring each other, creating broad coverage but limited additional technical differentiation.
Core inventive concept: The claims address the computational efficiency problem inherent in high-resolution autonomous vehicle camera feeds by reciting a specific mechanism: determining a field of view associated with a vanishing line indicative of a horizon line, generating a high-resolution crop portion corresponding to that field of view while downsampling the remainder of the image, and transmitting both to a neural network for simultaneous object detection — yielding long-range detection in the priority FOV without sacrificing the wider scene context from the downsampled portion.
Independent Claim Dissection
| Claim | Preamble | Transition | Key Body Elements |
|---|---|---|---|
| Claim 1 | A method | comprising: | determining, by a processor, a field of view of an image associated with a driving session wherein the field of view is associated with a vanishing line indicative of a horizon line; generating, by the processor, a crop portion corresponding to the field of view and a remaining portion wherein the remaining portion is downsampled; transmitting the crop portion and the downsampled remaining portion to a neural network causing detection of at least one object, wherein output is configured for use in autonomous drivingSearch prior art ↗ |
| Claim 8 | A system comprising one or more processors and non-transitory computer storage media storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations | comprising: | determining a field of view of an image associated with a driving session wherein the field of view is associated with a vanishing line indicative of a horizon line; generating a crop portion corresponding to the field of view and a remaining portion of the image wherein the remaining portion is downsampled; transmitting the crop portion and the downsampled remaining portion to a neural network causing detection of at least one object, wherein an output is configured for use in autonomous drivingSearch prior art ↗ |
| Claim 15 | A non-transitory computer storage media storing instructions that when executed by a system of one or more processors, cause the one or more processors to perform operations | comprising: | determining a field of view of an image associated with a driving session wherein the field of view is associated with a vanishing line indicative of a horizon line; generating a crop portion corresponding to the field of view and a remaining portion of the image wherein the remaining portion is downsampled; transmitting the crop portion and the downsampled remaining portion to a neural network causing detection of at least one object, wherein an output is configured for use in autonomous drivingSearch prior art ↗ |
Claim Dependency Tree
1 Method: determining FOV via vanishing/horizon line, generating crop + downsampled remainder, transmitting both to neural network for autonomous driving object detectionSearch Claim 1 prior art ↗
2 Adds: image is captured by a camera associated with a vehicleSearch in Eureka ↗
3 Adds: image is associated with a forward direction of a vehicleSearch in Eureka ↗
4 Adds: neural network receives additional data comprising at least data from an inertial measurement unitSearch in Eureka ↗
5 Adds: neural network receives additional data comprising at least map dataSearch in Eureka ↗
6 Adds: crop portion is aligned with the remaining portionSearch in Eureka ↗
7 Adds: neural network detects first object in crop portion and second object in remaining portion corresponding to same real-world object; method further comprises removing one of the first or second objectSearch in Eureka ↗
8 System (processor + non-transitory storage): same FOV/crop/downsample/neural network operations as Claim 1, output configured for autonomous drivingSearch Claim 8 prior art ↗
9 Adds: image is captured by a camera associated with a vehicleSearch in Eureka ↗
10 Adds: image is associated with a forward direction of a vehicleSearch in Eureka ↗
11 Adds: neural network receives additional data comprising at least data from an inertial measurement unitSearch in Eureka ↗
12 Adds: neural network receives additional data comprising at least map dataSearch in Eureka ↗
13 Adds: crop portion is aligned with the remaining portionSearch in Eureka ↗
14 Adds: neural network detects first object in crop and second in remaining portion corresponding to same real-world object; instruction causes processors to further remove one of the first or second objectSearch in Eureka ↗
15 CRM (non-transitory computer storage media): same FOV/crop/downsample/neural network operations as Claims 1 and 8, output configured for autonomous drivingSearch Claim 15 prior art ↗
16 Adds: image is captured by a camera associated with a vehicleSearch in Eureka ↗
| Metric | This Application | Software / AI / Automotive Norm |
|---|---|---|
| Total claims | 16 | 15 – 25 |
| Independent claim count | 3 | 2 – 4 |
| Dependent : Independent ratio | 4.33 : 1 | 5 – 8 : 1 |
| Method claims present? | Yes — Claim 1 | Always |
| System / apparatus claims? | Yes — Claim 8 | Always |
Analysis powered by PatSnap Eureka. Patent text and figures publicly available from USPTO. Draft a Similar Patent
Drafting Quality
Drafting Quality Signals
The patent demonstrates strong tripartite claim architecture with matching method (Claim 1), system (Claim 8), and CRM (Claim 15) independent claims, and the specification's detailed description provides robust technical grounding for the crop-and-downsample FOV mechanism. However, the dependent claim set suffers from near-identical mirroring across the three independent claims rather than introducing genuinely new technical fallback positions, and the vanishing line limitation in all three independent claims creates a potential design-around vector not addressed by any dependent claim.
✅
Antecedent Basis
All claims maintain clean antecedent basis throughout the claim set. In Claim 1, "the crop portion" and "the downsampled remaining portion" in the transmitting step correctly refer back to antecedents introduced in the generating step. In Claims 8 and 15, the same structural pattern is maintained, with "the image" referring back to the image introduced in the determining step. No orphaned "the" references were identified across all 16 claims.
✅
Spec–Claim Consistency
All three key limitations of the independent claims have direct specification support. The vanishing line / horizon line FOV determination maps to FIG. 4's "priority FOV" and "horizon" labels and the detailed description at column 4 lines 40–60. The crop-and-downsample generation step is fully illustrated in FIG. 2 (steps 204–208) and FIG. 4. The neural network transmission and detection step maps directly to FIG. 4's detector block (212) and the detailed description at columns 7–8. No independent claim limitation appears unsupported.
✅
Transition Word Usage
All three independent claims (1, 8, 15) use "comprising" as the transition word, which is the strategically optimal open-ended transition for this technology domain, preserving scope for additional unlisted system components and method steps. The specification's "Other Embodiments" section (column 13–14) explicitly defines "comprising" as inclusive and non-exclusive, reinforcing the broad interpretation. No missed opportunities for open-ended transition were identified.
✅
§112(f) Means-Plus-Function Risk
No "means for" or "step for" language appears in any of the 16 claims. The claim language uses active functional verbs with processor-based structural attribution ("determining, by a processor"; "generating, by the processor"; "transmitting, by the processor") throughout Claim 1, and system/CRM claims are tied to a concrete hardware recitation ("one or more processors and non-transitory computer storage media"). The risk of unintended §112(f) invocation through nonce words is low, though "neural network" could invite scrutiny as a black-box functional label if challenged.
⚠️
§101 Eligibility Risk
Under Alice step two, an examiner may characterize the core concept of Claims 1, 8, and 15 as the abstract idea of "processing images for object detection" — a well-litigated area. The hardware tie-in in Claim 8 ("one or more processors and non-transitory computer storage media") provides the strongest §101 defense, while Claim 1's method recitation is more vulnerable absent explicit hardware recitation. However, the "autonomous driving" output limitation and the vanishing line / horizon line determination add a practical application element. The specification's discussion of on-board processing and real-time autonomous vehicle operation (columns 3–4) provides prosecution arguments but is not mirrored in the independent claim language itself.
⚠️
Dependent Claim Fallback Quality
The dependent claims are structurally mirrored across all three independent claims — Claims 2–7 add the same limitations to Claim 1 as Claims 9–14 add to Claim 8, and Claim 16 adds only the camera limitation to Claim 15. This mirroring means the CRM independent claim (15) has only one dependent claim (16), creating a dangerously shallow fallback for the CRM branch. Claims 4/11 (IMU data) and 7/14 (duplicate object removal) add meaningful technical content, but no dependent claim addresses the specific CNN architecture (three-network pipeline of FIGS. 5A–5B), the batching mechanism, or the non-maximum suppression output combination — all of which are disclosed in the specification but unclaimed.
⚠️
Abstract Quality
An examiner reading only the abstract may identify the general crop-and-downsample mechanism but may conflate the invention with general image preprocessing techniques. The abstract accurately describes the hardware (image sensor positioned about a vehicle), the vanishing line association, and the CNN-based detection, but omits the specific "output configured for use in autonomous driving" limitation that provides the practical application anchor — the claim element most relevant to §101 eligibility. A stronger abstract would have foregrounded the autonomous driving use limitation and the specific efficiency gain (long-range faraway object detection without full-resolution computational cost).
✅
Figure Support Quality
Figure coverage is strong for the core claim elements. The FOV determination with horizon line is illustrated in FIG. 4 ("priority FOV" and "horizon" annotations). The crop generation is shown in FIG. 4 (step 206) and FIG. 2 (step 206). The downsampled remainder is shown in FIG. 4 (step 208 and the downsampled image path). The neural network detection is shown in FIG. 4 (detector block 212) and in full detail in FIGS. 5A and 5B. The duplicate removal step of Claim 7/14 is shown in FIG. 4 ("remove duplicates" arrow). No independent claim limitation lacks figure support, though the IMU and map data limitations of Claims 4/5/11/12 are only verbally described without dedicated figures.
Analysis powered by PatSnap Eureka. Patent text and figures publicly available from USPTO. Draft a Similar Patent
Scorecard
Strategic Intent Scorecard
Multi-dimensional assessment of this application's patent strategy quality, based on claim structure, specification depth, and prosecution positioning.
Claim Breadth
3.8
Prosecution Defensibility
3.5
Spec–Claim Consistency
4.2
Dependent Claim Coverage
2.8
Claim Type Diversity
4.5
Figure Support Quality
4
Key observation: Claim Type Diversity scores highest (4.5/5.0) because Tesla's filing successfully deploys all three standard claim formats — method (Claim 1), system (Claim 8), and CRM (Claim 15) — creating enforcement coverage against manufacturers, software developers, and end-system integrators simultaneously. Dependent Claim Coverage scores lowest (2.8/5.0) because the 13 dependent claims are almost entirely mirrored repetitions across the three independent claim branches rather than adding distinct technical fallback positions, leaving the three-CNN-network pipeline (FIGS. 5A–5B), batching mechanism, and NMS-based duplicate removal unclaimed in the dependent tier. Practitioners evaluating continuation or design-around strategies should note that the unclaimed CNN architecture and output-combination mechanism represent significant white space in the granted claim set.
See how your own draft compares — Open Eureka IP Drafting →
Critical Gaps
3 Critical Gaps in This Claim Set
A senior-attorney lens on the three highest-priority structural weaknesses — what each exposes in prosecution and litigation, and what a stronger filing would have done differently.
🔒
3 Critical Gaps in This Claim Set
See the full attorney-level analysis of what this application leaves unprotected — and how to draft it more defensively for your own filings.
CNN pipeline architecture not claimed Vanishing line mandatory design-around NMS output combination uncovered
Unlock Full Analysis — Free Frequently asked questions
US 12,198,396 B2 — key questions answered
Still have questions? PatSnap Eureka can answer them from patent data instantly. Search in Eureka
Disclaimer: This analysis is generated by PatSnap Eureka AI based on publicly available patent data from the USPTO. It does not constitute legal advice and should not be relied upon as such. Patent data may be subject to change as prosecution progresses. Scores and assessments reflect automated analysis and may not capture all relevant legal or technical nuances. Always consult a qualified patent attorney for formal legal opinions on patentability, freedom to operate, or infringement.
Ask anything about this patent.
PatSnap Eureka searches patents and data to answer instantly.
PatSnap Eureka searches patents and data to answer instantly.
Powered by PatSnap Eureka
Link copied to clipboard