DCT
4:20-cv-05342
Uniloc 2017 LLC v. Google LLC
I. Executive Summary and Procedural Information
- Parties & Counsel:
- Plaintiff: Uniloc 2017 LLC (Delaware)
- Defendant: Google LLC (Delaware)
- Plaintiff’s Counsel: Etheridge Law Group, PLLC
- Case Identification: 2:18-cv-00548, E.D. Tex., 12/30/2018
- Venue Allegations: Plaintiff alleges venue is proper in the Eastern District of Texas because Google maintains a "regular and established place of business" in the district. Allegations point to Google's physical Google Global Cache (GGC) servers hosted by ISPs in Tyler, Sherman, and Texarkana, as well as other services and facilities including Google Fi cellular service, Google Cloud Interconnect points of presence, and authorized repair centers.
- Core Dispute: Plaintiff alleges that Defendant’s Google Photos service infringes a patent related to a scalable method for content-based image retrieval from large databases.
- Technical Context: The lawsuit concerns content-based image retrieval (CBIR) technology, a field focused on searching and organizing large digital image libraries based on their visual content rather than manual text-based tags.
- Key Procedural History: The complaint alleges that Google has been on notice of the patent-in-suit since June 25, 2012, when it was cited during the prosecution of a Google patent application. This allegation may form the basis for a claim of willful infringement.
Case Timeline
| Date | Event |
|---|---|
| 1998-06-23 | ’201 Patent Priority Date |
| 2001-06-26 | ’201 Patent Issue Date |
| 2012-06-25 | Google allegedly put on notice of '201 Patent via office action |
| 2015-01-01 | Google Photos service launched (approximate date) |
| 2018-12-30 | Complaint Filing Date |
II. Technology and Patent(s)-in-Suit Analysis
U.S. Patent No. 6,253,201 - SCALABLE SOLUTION FOR IMAGE RETRIEVAL (Issued Jun. 26, 2001)
The Invention Explained
- Problem Addressed: The patent addresses the problem that comparing a target image to thousands or millions of other images in a large database is computationally intensive and slow, making real-time retrieval difficult (’201 Patent, col. 2:5-9). Traditional text-based keyword searching is also described as a burdensome and limited solution (’201 Patent, col. 1:11-29).
- The Patented Solution: The invention proposes a method to find similar images without direct image-to-image comparison. The system first partitions an image into a grid of content-independent blocks (’201 Patent, col. 3:30-32). It then analyzes the visual characteristics (e.g., color, edges) of each block to generate a corresponding "index value" (’201 Patent, Abstract). The system maintains lists of image identifiers that are associated with each index value for each partition location. To find a match for a target image, the system identifies the index values for the target's partitions, retrieves the corresponding lists of image identifiers, and accumulates a "count" for each image identifier that appears in those lists. Images with the highest counts are deemed the most similar (’201 Patent, Fig. 4).
- Technical Importance: This approach provided a method for making content-based image retrieval scalable, aiming to reduce the processing time required to search large, distributed image databases without a progressive degradation in performance as the database grows (’201 Patent, col. 2:10-18).
Key Claims at a Glance
- The complaint asserts at least independent claim 1 (Compl. ¶34).
- The essential elements of independent claim 1 include:
- partitioning a target image into a plurality of content-independent partitions,
- characterizing each partition of the plurality of content-independent partitions to form an index value associated with each partition,
- obtaining a list of image identifiers associated with the index value,
- accumulating counts of each image identifier in the list of image identifiers associated with each partition of the plurality of content-independent partitions, and
- retrieving at least one image associated with at least one of the image identifiers, based upon the counts of the at least one of the image identifiers.
- The complaint does not explicitly reserve the right to assert dependent claims.
III. The Accused Instrumentality
Product Identification
The accused product is the "Google Photos" service, including its mobile applications, website, and underlying cloud infrastructure (Compl. ¶17).
Functionality and Market Context
The complaint describes Google Photos as a service for organizing and searching photos, which are backed up on Google's cloud infrastructure (Compl. ¶17). A key accused feature is the ability to search for photos based on their content—such as "people, things, & places"—without requiring users to manually add tags (Compl. ¶¶18, 31). To accomplish this, the service allegedly uses machine learning frameworks like TensorFlow and models such as the Single Shot Multibox Detector (SSD) to analyze, characterize, and index images for later retrieval (Compl. ¶¶22, 29). The complaint highlights the service's popularity, noting over a billion installs from the Google PlayStore (Compl. ¶17). An image from Google's marketing materials shows a user searching for "dog" and retrieving corresponding photos. (Compl. p. 31, "Find your photos faster").
IV. Analysis of Infringement Allegations
’201 Patent Infringement Allegations
| Claim Element (from Independent Claim 1) | Alleged Infringing Functionality | Complaint Citation | Patent Citation |
|---|---|---|---|
| partitioning a target image into a plurality of content-independent partitions, | Google Photos allegedly uses a Single Shot Detector (SSD) algorithm that partitions an image into a set of "bounding boxes (partitions)" at different scales and aspect ratios. The complaint alleges these partitions are "content independent" (Compl. ¶27). | ¶¶26-27 | col. 8:2-4 |
| characterizing each partition ... to form an index value associated with each partition, | The SSD algorithm allegedly "characterizes each bounding box (or, partition) for predicting classes (or, index values)" and generates "labels/ classes (or index values) e.g. object names" such as "bowl, cup, chair, etc." (Compl. ¶29). | ¶¶28-29 | col. 8:5-7 |
| obtaining a list of image identifiers associated with the index value, | The generated labels/classes are allegedly associated with image identifiers through an indexing process. The complaint states that "document identifiers are appended/added to list of documents associated with specific index values" (Compl. ¶32). A screenshot shows a map of GGC edge nodes. (Compl. p. 8). | ¶¶32-33 | col. 8:8-9 |
| accumulating counts of each image identifier in the list of image identifiers..., and | The complaint alleges direct infringement of the claim but does not specify how counts are accumulated. It states the system retrieves images based on user queries against the indexed labels, which are used for retrieving images (Compl. ¶31). | ¶¶31, 34 | col. 8:10-13 |
| retrieving at least one image ... based upon the counts of the at least one of the image identifiers. | Google Photos allegedly retrieves images that match a user's query (e.g., a search for "dog"). A graphic from the Google Photos website illustrates this search and retrieval functionality (Compl. p. 36, "Find your photos faster"). | ¶¶30-31, 34 | col. 8:14-18 |
Identified Points of Contention
- Scope Questions: A primary question may be whether the "bounding boxes" generated by the accused SSD algorithm (Compl. ¶27) meet the claim limitation of "content-independent partitions". While the complaint asserts they are content independent, the patent's specification describes partitions as a more structured, grid-like array (e.g., "4x4, 8x8, or 16x16 partitioning") (’201 Patent, col. 3:30-32). The court may need to determine if the claim scope covers the accused dynamic, variably-sized partitions.
- Technical Questions: Claim 1 requires the step of "accumulating counts" of image identifiers to facilitate retrieval. The complaint alleges infringement of the claim as a whole but focuses its technical description on the indexing of labels and retrieval via search queries (Compl. ¶¶31-33). A key evidentiary question will be whether the accused Google Photos system performs a function that is structurally or functionally equivalent to "accumulating counts," or if its retrieval algorithm operates in a fundamentally different way that falls outside the claim.
V. Key Claim Terms for Construction
"content-independent partitions"
- Context and Importance: This term is central to the infringement analysis. The patent's inventive concept appears to rely on a pre-defined, structured division of an image, whereas the accused product allegedly uses a machine learning algorithm to generate "bounding boxes." Practitioners may focus on this term because its construction could determine whether the accused SSD algorithm's method of dividing an image falls within the scope of the claim.
- Intrinsic Evidence for Interpretation:
- Evidence for a Broader Interpretation: The claim language itself does not specify the shape, size, or method of creating the partitions, which may support an interpretation that any method of dividing an image into predefined regions, regardless of how those regions are defined, is covered.
- Evidence for a Narrower Interpretation: The specification repeatedly provides examples of fixed, grid-like arrays, stating "typically the array is a 4x4, 8x8, or 16x16 partitioning of the image" (’201 Patent, col. 3:30-32). This could support an argument that the term requires a static, grid-based structure, potentially excluding the variably-sized "bounding boxes" generated by the accused SSD algorithm.
"accumulating counts"
- Context and Importance: This phrase defines the specific mechanism for scoring similarity in the patent. The complaint's description of the accused system's retrieval logic is less specific on this point. The viability of the infringement claim may depend on whether Google Photos' search algorithm can be shown to perform this step, either literally or equivalently.
- Intrinsic Evidence for Interpretation:
- Evidence for a Broader Interpretation: A party might argue that any aggregation of similarity signals from different partitions to generate a final relevance score constitutes "accumulating counts" in a functional sense.
- Evidence for a Narrower Interpretation: The patent's flowchart (FIG. 4, step 444) and description point to a more literal summation: "the count of each image identifier ID that is contained in the extracted list ... is accumulated" (’201 Patent, col. 6:8-10). This may support a narrower construction requiring a direct tallying process, which could differ from a more complex relevance-scoring or machine-learning-based ranking system.
VI. Other Allegations
Indirect Infringement
The complaint alleges Google induces infringement by providing customers with instructions, demonstrations, and user guides via its websites that allegedly instruct them to use Google Photos in an infringing manner (Compl. ¶36).
Willful Infringement
The complaint alleges pre-suit knowledge based on the assertion that Google was on notice of the ’201 Patent as of June 25, 2012, because the patent was cited by the USPTO in a non-final rejection of a Google patent application (Compl. ¶38). Willfulness is also alleged based on post-suit knowledge via service of the complaint itself (Compl. ¶38).
VII. Analyst’s Conclusion: Key Questions for the Case
- A core issue will be one of definitional scope: can the claim term "content-independent partitions," which the patent specification exemplifies as fixed grids, be construed to read on the dynamically generated, variably-sized "bounding boxes" allegedly created by Google Photos' machine learning algorithm?
- A key evidentiary question will be one of functional operation: does the accused Google Photos search algorithm perform the specific claimed step of "accumulating counts" to determine image similarity, or does it employ a different ranking and retrieval logic that creates a fundamental mismatch with the patented method?
- A third significant issue relates to willfulness: if infringement is found, the court will need to consider whether Google's alleged pre-suit knowledge of the patent, stemming from a 2012 patent office action, is sufficient to establish the egregious conduct necessary to support an enhancement of damages.